From WikiContent

Revision as of 22:27, 16 September 2006

Groupchat, : Components, and Event Models

By now, you should have a

good idea of how scripts interact with Jabber and how the core

elements such as <message/> and

<presence/> can be constructed and handled.

This chapter builds upon what we've already seen in Chapter 8 and introduces new concepts. We build a nosy assistant that joins a conference room and alerts us to words and phrases that we want it to listen for. There are two popular conference protocols, as mentioned in Section 6.2.6—the presence-based Groupchat protocol, and the jabber:iq:conference-based Conference protocol. The assistant recipe, a foray into the world of 'bots, takes a look at the original presence-based one.

As we've seen, programming within Jabber's event model is fairly straightforward. But what happens when you want to meld other components with event models of their own? We look at a couple of typical scenarios where this melding needs to happen. The first is a homage to the Trojan Room Coffee Machine (http://www.cl.cam.ac.uk/coffee/coffee.html), where we give life, or at least presence, to a coffeepot, using LEGO® MINDSTORMS™. The second is a Tk-based RSS headline viewer. Both the coffeepot and the Tk programming library have event loops of their own. With the coffeepot, we need to have a loop that polls the coffeepot's status, independently of the polling for incoming packets from the Jabber server. The Tk programming library's event model is similar to those of the Jabber programming libraries that are used in the recipes in this book, in that handlers are set up and a loop is started that listens for UI events. In both cases, we need to get these event loops to work in harmony with the Jabber libraries' event loops.

Building Jabber solutions without event loops is a Sisyphean task. The very nature of Jabber communication is event-based, and it's important to understand how to use the event features of the Jabber programming libraries and also how to be able to mesh those features with similar features in other libraries and systems.

We also look at extending messages and build a mechanism that delivers RSS headlines to clients who register with that mechanism. These headlines are carried using an extended message type. In fact, the RSS Delivery Mechanism is a component. The three recipes in Chapter 8, were Jabber clients, in that they connected to the Jabber network via the Jabber Session Manager (JSM) service. We look at the differences between programming a client and programming a component in this chapter and build a complete component that can be queried and interacted with using the third of Jabber's building blocks—the <iq/> element.

Happy coding!

Keyword Assistant

Many of the Jabber core and peripheral developers hang out in a conference room called jdev hosted by the Conferencing component on the Jabber server running at jabber.org. While a lot of useful information is to be gleaned from listening to what goes on in jdev, it isn't possible to be there all the time. Conversations in jdev are logged to web pages, which can be used to visit after the fact to try to catch up with things; however, this can be a hopeless task. One solution is to build a 'bot that looks for specific keywords and Uniform Resource Locators (URLs) in the conversations in jdev and send those on as Jabber messages.

This script, keyassist, connects to a Jabber server, enters a conference room, and listens to the conversations, looking for certain words and phrases to be uttered. The keyassist script is given a bit of "intelligence" in that it can be interacted with and told, while running, to watch for (or stop watching for), certain words and phrases.

The keyassist script introduces us to programmatic interaction with the Conferencing component. Before looking at the script, however, let's have a brief overview of Conferencing in general.

Conferencing

The Conferencing component at jabber.org is conference.jabber.org. Details of the component instance configuration for such a Conferencing component can be found in Section 4.10, where we see that the component exists as a shared object library connected with the library load component connection method. This component provides general conferencing facilities, oriented around a conference room and conference user model.

A Jabber user can enter (or join) a conference room, thereby becoming a conference user identified by a nickname that is chosen upon entering that room. Nicknames are generally used in conference rooms to provide a modicum of privacy—it is assumed that by default you don't want to let the other conference room members know your real JID.

The Conferencing component supports two protocols for user and room interaction: a simple one that provides basic features and a more complex one that provides the basic features plus facilities such as password-protected rooms and room descriptions—Groupchat and Conference.

{{Note|There is a third protocol, called Experimentaliq:groupchat, which came between the Groupchat and Conference protocols. This reflected an experimental move to add features to the basic Groupchat protocol using IQ elements, the contents of which were qualified by a namespace jabber:iq:groupchat. This protocol has been dropped, and support for it exists only in certain versions of WinJab and JIM.

</code>

Groupchat

The Groupchat protocol is the simpler of the two and provides basic

functions for entering and exiting conference rooms and choosing

nicknames. : This Groupchat protocol is known as the presence-based

protocol, because the protocol is based upon

<presence/> elements used for room entry, exit, and

nickname determination. The Groupchat protocol has a nominal version

number of 1.0.

Conference

The Conference protocol offers more advanced features than the

Groupchat protocol and makes use of two IQ namespaces:

jabber:iq:conference and jabber:iq:browse. It has a

nominal protocol version number of 1.4, which reflects the version of

the Jabber server with which it is delivered. Sometimes this version

number is referred to as 0.4, such as in the downloadable tarball and

in the value returned in response to a "version query" on the

component itself, as shown in Example 9-1. : The version number isn't

that important. The main thing to keep in mind is that the component

that is called conference.so (see the reference to the

shared object library in Section 4.10.4) supports both the

Groupchat protocol and the Conference protocol. If you come across a

shared object library called groupchat.so, this is the

original Conferencing component that was made available with Jabber

server Version 1.0. This library supports only the Groupchat protocol.

Querying the Conferencing component's version

SEND: <iq type='get' to='conference.gnu.mine.nu'> <query xmlns='jabber:iq:version'/> </iq>

RECV: <iq to='dj@gnu.mine.nu/jarl' from='conference.gnu.mine.nu' type='result'> <query xmlns='jabber:iq:version'> <name>conference</name> <version>0.4</version> <os>Linux 2.4.2-2</os> </query> </iq> In this recipe we'll be using the simpler Groupchat protocol. It's widely used and easy to understand. Example 9-2 shows a typical element log from Groupchat-based activity. It shows a user, with the JID qmacro@jabber.com, entering a room called "cellar," hosted on the conference component at conf.merlix.dyndns.org, a room that currently has two other occupants who go by the nicknames flash and roscoe. The elements are from qmacro's perspective, and are all explained following the example.

The Groupchat protocol in action The user qmacro tries to enter the conference room with the nickname flash and fails:

SEND: <presence to='cellar@conf.merlix.dyndns.org/flash'/>

RECV: <presence to='qmacro@jabber.com/jarltk' from='cellar@conf.merlix.dyndns.org/flash' type='error'> <error code='409'>Conflict</error> </presence>

He tries again, this time with a different nickname, deejay, and is successful:

SEND: <presence to='cellar@conf.merlix.dyndns.org/deejay'/>

RECV: <presence to='qmacro@jabber.com/jarltk' from='cellar@conf.merlix.dyndns.org/flash'/>

RECV: <presence to='qmacro@jabber.com/jarltk' from='cellar@conf.merlix.dyndns.org/roscoe'/>

RECV: <presence to='qmacro@jabber.com/jarltk' from='cellar@conf.merlix.dyndns.org/deejay'/>

RECV: <message to='qmacro@jabber.com/jarltk' type='groupchat' from='cellar@conf.merlix.dyndns.org'> <body>deejay has become available</body> </message>

roscoe says hi, and qmacro waves back:

RECV: <message to='qmacro@jabber.com/jarltk' from='cellar@conf.merlix.dyndns.org/roscoe' type='groupchat' cnu=> <body>hi</body> </message>

SEND: <message to='cellar@conf.merlix.dyndns.org' type='groupchat'> <body>/me waves to everyone</body> </message>

flash sends a private message to qmacro:

RECV: <message to='qmacro@jabber.com/jarltk' from='cellar@conf.merlix.dyndns.org/flash' type='chat'> <body>Is that you, qmacro?</body> <thread>jarl1998911094</thread> </message>

Feeling left out of the conversation, roscoe leaves the room:

RECV: <presence to='qmacro@jabber.com/jarltk' type='unavailable' from='cellar@conf.merlix.dyndns.org/roscoe'/>

RECV: <message to='qmacro@jabber.com/jarltk' type='groupchat' from='cellar@conf.merlix.dyndns.org'> <body>roscoe has left</body> </message> Let's take the stages in Example 9-2 one by one.

Failed attempt to enter room

qmacro makes an attempt to enter the room using the Groupchat

protocol. This is done by sending a directed

<presence/> element to a particular JID that represents

the room and the chosen nickname. This JID is constructed as follows:

[room name]@[conference component]/[nickname]

In this example, the conferencing component is identified with the

hostname conf.merlix.dyndns.org. qmacro's choice of

nickname is flash:

cellar@conf.merlix.dyndns.org/flash

Thus the following element is sent:

SEND: <presence to='cellar@conf.merlix.dyndns.org/flash'/>

The conference component determines that there is already someone

present in the room cellar@conf.merlix.dyndns.org with the

nickname flash, so qmacro is notified of this and

receives a directed presence with an <error/> tag:

RECV: <presence to='qmacro@jabber.com/jarltk' from='cellar@conf.merlix.dyndns.org/flash' type='error'> <error code='409'>Conflict</error> </presence>

Note that the <presence/> element has the type

error and comes from the artificial JID constructed in the

room entry attempt. The element is addressed to qmacro's real

JID, of course—qmacro@jabber.com/jarltk—as otherwise it

wouldn't reach him. : The error code 409 and text "Conflict" tells

qmacro that the nickname conflicted with one already in the

room. This is a standard error code/text pair; Table 5-3 shows a

complete set of code/text pairs. : At this stage, qmacro is

not yet in the room.

Successful attempt to enter room

qmacro tries again, this time with a different nickname,

deejay:^[1]

SEND: <presence to='cellar@conf.merlix.dyndns.org/deejay'/>

This time, there is no conflict—no other user is in the room "cellar"

with that nickname—and the conference component registers the entry.

It does this by sending qmacro the presence of all the room

occupants, including that of himself:

RECV: <presence to='qmacro@jabber.com/jarltk' from='cellar@conf.merlix.dyndns.org/flash'/>

RECV: <presence to='qmacro@jabber.com/jarltk' from='cellar@conf.merlix.dyndns.org/roscoe'/>

RECV: <presence to='qmacro@jabber.com/jarltk' from='cellar@conf.merlix.dyndns.org/deejay'/>

These presence elements are also sent to the other room occupants so

they know that deejay is present.

Conference component-generated notification

In addition to the presence elements sent for each room occupant, a

general roomwide message noting that someone with the nickname

deejay just entered the room is sent out by the component as

a type='groupchat' message to all the room occupants:

RECV: <message to='qmacro@jabber.com/jarltk' type='groupchat' from='cellar@conf.merlix.dyndns.org'> <body>deejay has become available</body> </message>

The text "has become available" used in the body of the message is

taken directly from the Action Notices definitions, part of the

Conferencing component instance configuration described in Section

4.10.3. Note that the identity of the room itself is simply a generic

version of the JID that the room occupants use to enter:

cellar@conf.merlix.dyndns.org

Roomwide chat

Once the user with the nickname roscoe sees someone enter the

room, he sends a greeting, and qmacro waves back:

RECV: <message to='qmacro@jabber.com/jarltk' from='cellar@conf.merlix.dyndns.org/roscoe' type='groupchat' cnu=> <body>hi qmacro</body> </message>

SEND: <message to='cellar@conf.merlix.dyndns.org' type='groupchat'> <body>/me waves to everyone</body> </message>

As with the notification message, each message is a

groupchat-type message. The one received appears to come from

cellar@conf.merlix.dyndns.org/roscoe, which is the JID

representing the user in the room with the nickname roscoe.

This way, roscoe's real JID is never sent to qmacro.

The message deejay sends is addressed to the room's identity

cellar@conf.merlix.dyndns.org, and contains a message that

starts with /me. This is simply a convention that is

understood by clients that support conferencing, meant to represent an

action and displayed thus:

* deejay waves to everyone

{{Note|Ignore the cnu attribute; it's put there and used by the component and should never make it out to the client endpoints. The attribute name is a short name for the conference user and refers to the internal structure that represents a conference room occupant within the component.

</code>

One-on-one chat

The Conferencing component also supports a one-on-one chat mode, which

is just like normal chat mode (where messages with the type

chat are exchanged) except that the routing goes through the

component. The intended recipient of a conference-routed chat message

is identified by his room JID. So in this example:

RECV: <message to='qmacro@jabber.com/jarltk' from='cellar@conf.merlix.dyndns.org/flash' type='chat'> <body>Is that you, qmacro?</body> <thread>jarl1998911094</thread> </message>

the user nicknamed flash actually addressed the chat message

to the JID:

cellar@conf.merlix.dyndns.org/deejay

which arrived at the Conferencing component (because of the hostname,

conf.merlix.dyndns.org causes the <message/>

element to be routed there), which then looked up internally who

deejay really was (qmacro@jabber.com/jarltk) and

sent it on. This way, the recipient of a conference-routed message

never discovers the real JID of the sender. In all other ways, the

actual <message/> element is like any other

<message/> element—in this case, it contains a message

<body/> and a chat <thread/>. (See

Section 5.4.1 for details on the <message/> element.)

Leaving the room

In the same way that room entrance is effected by sending an

available presence (remember, a <presence/>

element without an explicit type attribute is understood to

represent type="available'), leaving a room is achieved by

doing the opposite:

RECV: <presence to='qmacro@jabber.com/jarltk' type='unavailable' from='cellar@conf.merlix.dyndns.org/roscoe'/>

The people in the conference room are sent a message that

roscoe has left the room by the unavailable presence

packet. This is by and large for the benefit of each user's client, so

that the room occupant list can be updated. The component also sends

out a verbal notification, in the same way as it sends a verbal

notification out when someone joins:

RECV: <message to='qmacro@jabber.com/jarltk' type='groupchat' from='cellar@conf.merlix.dyndns.org'> <body>roscoe has left</body> </message>

Like the join notification, the text for the leave notification ("has

left") comes directly from the component instance configuration

described in Section 4.10.3.

The Script's Scope

The Keyword Assistant (keyassist) script will be written in Python using the Jabberpy library. As mentioned earlier, the script will perform the following tasks:

• Connect to a predetermined Jabber server
• Join a predetermined

conference room

• Sit there quietly, listening to the conversation
• Take simple commands from people to watch for, or stop watching for,

particular words or phrases uttered in the room

• Relay the context of those words or phrases to whomever requested them, if heard In addition

to setting the identity of the Jabber server and the conference room in variables, we'll also need to keep track of which users ask the assistant for words and phrases.
We'll use a dictionary (hash in Perl terms), as shown in Example 9-3, because we want to manage the data in there by key, the JID of those users that the script will be assisting. Having a look at what this dictionary will look like during the lifetime of this script will help us to visualize what we're trying to achieve.

Typical contents of the Keyword Assistant's dictionary

{ 'dj@gnu.pipetree.com/home': { 'http:': 1, 'ftp:': 1

                                         },

 'piers@jabber.org/work':                { 'Perl': 1, 'Java': 1, 'SAP
 R/3': 1
                                         },

 'cellar@conf.merlix.dyndns.org/roscoe': { 'dialback': 1
                                         }

We can see from the contents of the dictionary in Example 9-3 that three people have asked the script to look out for words and phrases. Two of those people—dj and piers—have interacted with the script directly by sending a normal (or chat) <message/>. The other person, with the conference nickname roscoe, is in the "cellar" room and has sent the script a message routed through the Conference component in the same way that flash sent qmacro a private message in Example 9-2: the JID of the sender belongs to (has the hostname set to) the conference component. Technically, there's nothing to distinguish the three JIDs here; it's just that we know from the name that conf.merlix.dyndns.org is the name that identifies such a component.

If we dissect the dictionary, we can see that:

• dj wants to be notified if any web or FTP URLs are mentioned.
• piers is interested in references to two of his favorite

languages and his favorite business software solution.

• roscoe

is interested in any talk about dialback. We said we'd give the script a little bit of intelligence. This was a reference to the ability for users to interact with the script while it runs, rather than having to give the script a static list of words and phrases in a configuration file. dj, piers, and roscoe have done this by sending the script messages (directly, not within the room) with simple keyword commands, such as:

dj: "watch http:" script: "ok, watching for http:" dj:

"watch gopher:" script: "ok, watching for gopher:" dj:

"watch ftp:" script: "ok, watching for ftp:" dj: "ignore

gopher:" script: "ok, now ignoring gopher:"

...

piers: "list" script: "watching for: Perl, Java, SAP R/3"

...

roscoe: "stop" script: "ok, I've stopped watching"

The keyassist Script

Example 9-4 shows the keyassist script in its entirety. The script is described in detail, step by step, in the next section.

The keyassist Perl script

import jabber from string import split, join, find import sys

keywords = {}

def addword(jid, word): if not keywords.has_key(jid): keywords[jid] = {} keywords[jid][word] = 1

def delword(jid, word): if keywords.has_key(jid): if keywords[jid].has_key(word): del keywords[jid][word]

def messageCB(con, msg):

   type = msg.getType() if type == None: type = 'normal'

   # Deal with interaction
   if type == 'chat' or type == 'normal': jid = str(msg.getFrom())

       message = split(msg.getBody(), None, 1); reply = ""

       if message[0] == 'watch': addword(jid, message[1]) reply =
       "Okay, watching for " + message[1]

       if message[0] == 'ignore': delword(jid, message[1]) reply =
       "Okay, now ignoring " + message[1]

       if message[0] == 'list': if keywords.has_key(jid): reply =
       "Watching for: " + join(keywords[jid].keys(), ", ") else: reply
       = "Not watching for any keywords"

       if message[0] == 'stop': if keywords.has_key(jid): del
       keywords[jid] reply = "Okay, I've stopped watching"

       if reply: con.send(msg.build_reply(reply))

   # Scan room talk
   if type == 'groupchat': message = msg.getBody()

       for jid in keywords.keys(): for word in keywords[jid].keys(): if
       find(message, word) >= 0: con.send(jabber.Message(jid, word +
       ": " + message))

   # Deal with nickname conflict in room
   if str(prs.getFrom()) == roomjid and prs.getType() == 'error':
   prsnode = prs.asNode() error = prsnode.getTag('error') if error: if
   (error.getAttr('code') == '409'): print "Cannot join room -
   conflicting nickname" con.disconnect() sys.exit(0)

   # Remove keyword list for groupchat correspondent
   if prs.getType() == 'unavailable': jid = str(prs.getFrom()) if
   keywords.has_key(jid): del keywords[jid]

Dissecting the keyassist Script

Taking keyassist step by step, the first section is probably familiar if you've seen the previous Python-based scripts in Section 8.1 and Section 8.3, both in Chapter 8.

import jabber from string import split, join, find import sys

Here, all of the functions and libraries that we'll need are brought in. We'll use the find function from the string library to help with the keyword searching.

Next, we declare the dictionary. This will hold a list of the words that the script will look for, as defined by each person, as shown in Example 9-3.

keywords = {}

Maintaining the keyword dictionary

To maintain this dictionary, we will use two subroutines to add words to and remove words from a user's word list. These subroutines are called when a command such as watch or ignore is recognized in the callback subroutine that handles incoming <message/> elements:

def addword(jid, word): if not keywords.has_key(jid): keywords[jid] = {} keywords[jid][word] = 1

def delword(jid, word): if keywords.has_key(jid): if keywords[jid].has_key(word): del keywords[jid][word]

A string representation of the JID (in jid) of the correspondent giving the command is passed to the subroutines along with the word or phrase specified (in word) by the user. The dictionary has two levels: the first level is keyed by the JID, and the second by word or phrase. We use a dictionary, rather than an array, at the second level simply to make removal of words and phrases easier.

Message callback

Next, we define the callback to handle incoming <message/> elements:

def messageCB(con, msg):

   type = msg.getType() if type == None: type = 'normal'

As usual, we're expecting the message callback to be passed the connection object (in con) and the message object itself (msg). How this callback is to proceed is determined by the type of message received. We determine the type (taken from the <message/> element's type attribute) and store it in the variable called type. Remember that if no type attribute is present, a message type of normal is assumed. (See Section 5.4.1.1 for details of <message/> attributes.)

The two types of incoming messages we're expecting this script to receive are those conveying the room's conversation—in groupchat-type messages—and those over which the commands such as watch and ignore are carried, which we expect in the form of normal- or chat-type messages.

The first main section of the messageCB handler deals with incoming commands:

# Deal with interaction if type == 'chat' or type == 'normal': jid = str(msg.getFrom())

       message = split(msg.getBody(), None, 1); reply = ""

       if message[0] == 'watch': addword(jid, message[1]) reply =
       "Okay, watching for " + message[1]

       if message[0] == 'ignore': delword(jid, message[1]) reply =
       "Okay, now ignoring " + message[1]

       if message[0] == 'list': if keywords.has_key(jid): reply =
       "Watching for: " + join(keywords[jid].keys(), ", ") else: reply
       = "Not watching for any keywords"

       if message[0] == 'stop': if keywords.has_key(jid): del
       keywords[jid] reply = "Okay, I've stopped watching"

       if reply: con.send(msg.build_reply(reply))


If the <message/> element turns out to be of the type in which we're expecting a potential command, we want to determine the JID of the correspondent who sent that message. Calling the getFrom() method will return us a JID object. What we need is the string representation of that, which can be determined by calling the str() function on that JID object:

jid = str(msg.getFrom())

Then we grab the content of the message by calling the getBody() on the msg object and split the whole thing on the first bit of whitespace. This should be enough for us to distinguish a command (watch, ignore, and so on) from the keywords. After the split, the first element (index 0) in the message array will be the command, and the second element (index 1) will be the word or phrase, if given. At this stage, we also declare an empty reply:

message = split(msg.getBody(), None, 1); reply = ""

Now it's time to determine if what the script was sent made sense as a command:

if message[0] == 'watch': addword(jid, message[1]) reply = "Okay, watching for " + message[1]

       if message[0] == 'ignore': delword(jid, message[1]) reply =
       "Okay, now ignoring " + message[1]

       if message[0] == 'list': if keywords.has_key(jid): reply =
       "Watching for: " + join(keywords[jid].keys(), ", ") else: reply
       = "Not watching for any keywords"

       if message[0] == 'stop': if keywords.has_key(jid): del
       keywords[jid] reply = "Okay, I've stopped watching"

We go through a series of checks, taking appropriate action for the supported commands:

watch

Watch for a particular word or phrase.

ignore

Stop watching for a particular word or phrase.

list

List the words and phrases currently being watched.

stop

Stop watching altogether; remove the list of words and phrases. The

addword() and delword() functions defined earlier

are used here, as well as other simpler functions; one that lists the

words and phrases for a particular JID:

keywords[jid].keys()

and one that removes them:

del keywords[jid]

If there was something recognizable for the script to do, we get it to reply appropriately:

if reply: con.send(msg.build_reply(reply))

The build_reply() function creates a reply out of a message object by setting to to the value of the original <message/> element's from attribute and preserving the element type attribute and <thread/> tag, if present. The <body/> of the reply object (which is just a <message/> element) is set to whatever is passed in the function call; in this case, it's the text in the reply variable.

Now that we've dealt with incoming commands, we need another section in the message callback subroutine to scan for the words and phrases. The target texts for this scanning will be the snippets of room conversation, which arrive at the callback in the form of groupchat-type <message/> elements:

# scan room talk if type == 'groupchat': message = msg.getBody()

The message variable holds the string we need to scan; it's just a case of checking for each of the words or phrases on behalf of each of the users who have asked:

for jid in keywords.keys(): for word in keywords[jid].keys(): if find(message, word) >= 0: con.send(jabber.Message(jid, word + ": " + message))

If we get a hit, we construct a new Message object, passing the JID of the person for whom the string has matched (in the jid variable) and the notification, consisting of the word or phrase that was found (in word) and the context in which it was found (the sentence uttered, in message). Once found and constructed, the <message/> is sent to that user. By default, the Message constructor specifies no type attribute, so the user is sent a "normal" message.

Presence callback

Having dealt with the incoming <message/> elements, we turn to the <presence/> elements. Most of those we receive in this conference room will be notifications from people entering and leaving the room, as shown in Example 9-2. We want to perform housekeeping on our keywords dictionary so the entries don't become stale. We also want to deal with the potential problem of conflicting nicknames. Let's look at that first.

We want to check for the possibility of nickname conflict problems that may occur when we enter the room, and the chosen nickname (flash) is already taken.

Remembering that a conflict notification will look something like this:

<presence to='qmacro@jabber.com/jarltk' from='cellar@conf.merlix.dyndns.org/flash' type='error'> <error code='409'>Conflict</error> </presence> we test for the receipt of a <presence/> element with the following:

def presenceCB(con, prs):

   # Deal with nickname conflict in room
   if str(prs.getFrom()) == roomjid and prs.getType() == 'error':
   prsnode = prs.asNode() error = prsnode.getTag('error') if error: if
   (error.getAttr('code') == '409'): print "Cannot join room -
   conflicting nickname" con.disconnect() sys.exit(0)

The <presence/> element will appear to be sent from the JID that we constructed for the initial room entry negotiation (in the roomjid variable further down in the script); for example:

jdev@conference.jabber.org/kassist We compare this value to the value of the incoming <presence/>'s from attribute, and also make sure that the type attribute is set to error. If it is, we want to extract the details from the <error/> tag that will be contained as a direct child of the <presence/>.

The Jabberpy library currently doesn't offer a direct high-level function to get at this tag from the Presence object (in prs), but we can strip away the presence object "mantle" and get at the underlying object, which is a neutral "node"—a Jabber element, or XML fragment, without any preconceived ideas of what it is (and therefore without any accompanying high-level methods such as getBody() or setPriority()).

{{Note|If this seems a little cryptic, just think of it like this: each of the Presence, Message, and IQ classes are merely superclasses of the base class Protocol, which represents elements generically.

</code> The asNode() method gives us what we need—a Protocol object representation of the <presence/> element. From this we can get to the <error/> tag and its contents. If we find that we do have a nickname conflict, we abort by disconnecting from the Jabber server and ending the script.

The general idea is that this script will run indefinitely and notify the users on a continuous basis, so we need to do a spot of keyword housekeeping. No presence subscription relationships are built (mostly to keep the script small and simple; you could adapt the mechanism from the recipe in Section 8.3 if you wanted to make this script sensitive to presence), so notifications will get queued up for the user if he is offline with the use of the mod_offline module of the Jabber Session Manager (JSM). This makes a lot of sense for the most part; however, we still want to have the script send notifications even if the user is offline. Additionally, a command could be sent to the script to watch for a keyword or phrase from a user within the room. We would receive the command from a JID like this:

jdev@conference.jabber.org/nickname This is a transient JID, in that it represents a user's presence in the jdev room for a particular session. If a word is spotted by the script hours or days later, there's a good chance that the user has left the room, making the JID invalid as a recipient. Although the JID is technically valid and will reach the conferencing component, there will be no real user JID that it is paired up with. Potentially worse, the room occupant's identity JID may be assigned to someone else at a later stage, if the original user left, and a new user entered choosing the same nickname the original user had chosen. See the upcoming sidebar titled "Transient and Nonexistent JIDs" for a short discussion of the difference between a transient JID and a nonexistent JID.

So as soon as we notice a user leave the room we're in, which will be indicated through a <presence/> element conveying that occupant's unavailability, we should remove any watched-for words and phrases from the dictionary:

# Remove keyword list for groupchat correspondent if prs.getType() == 'unavailable': jid = str(prs.getFrom()) if keywords.has_key(jid): del keywords[jid]

As before, we obtain the string representation of the JID using the str() function on the JID object that represents the presence element's sender, obtained via the getFrom() method.

{{Sidebar|Transient and Nonexistent JIDsWhat happens when you send a message to a "transient" conference room JID? Superficially, the same as when you send one to a nonexistent JID. But there are some subtle differences.

A transient JID is one that reflects a user's alternate identity in the context of the Conferencing component. When you construct and send a message to a conference transient JID, it goes first to the conference component because of the hostname in the JID that identifies that component, for example:

jdev@conference.jabber.org/qmacro The hostname conference.jabber.org is what the jabberd backbone uses to route the element. As mentioned earlier, the Conferencing component will relay a message to the real JID that belongs to the user currently in a room hosted by that component.

While the component itself is usually persistent, the room occupants (and so their transient JIDs) are not. When a message is sent to the JID jdev@conference.jabber.org/qmacro and there is no room occupant in the jdev room with the nickname qmacro, the message will still reach its first destination—the component—but be rejected at that stage, as shown in Example 9-5.

A message to a nonexistent transient JID is rejected

SEND: <message to='jdev@conference.jabber.org/qmacro'> <body>Hello there</body> </message>

RECV: <message to='dj@gnu.mine.nu/jarl' from='jdev@conference.jabber.org/qmacro' type='error'> <body>Hello there</body> <error code='404'>Not Found</error> </message>

Although the rejection—the "Not Found" error—is the same as if a message had been sent to a JSM user that didn't exist, the difference is that the transient user always had the potential to exist, whereas the JSM user never did. Of course, if the JID referred to a nonexistent Jabber server, then the error returned wouldn't be a "Not Found" error 404, but an "Unable to resolve hostname" error 502.

</code>

The main script

Now that we have the subroutines and callbacks set up, all we need to do is define the Jabber server and room information:

Server = 'gnu.mine.nu' Username = 'kassist' Password = 'pass' Resource = 'py'

Room = 'jdev' ConfServ = 'conference.jabber.org' Nick = 'kassist'

The kassist user can be set up simply by using the reguser script presented in Section 7.4:

$ ./reguser gnu.mine.nu username=kassist password=pass [Attempt] (kassist) Successful registration $ In the same way as in previous recipes' scripts, a connection attempt is made, followed by an authentication attempt:

con = jabber.Client(host=Server,debug=0,log=0) try: con.connect() except IOError, e: print "Couldn't connect: %s" % e sys.exit(0) else: print "Connected"

if con.auth(Username,Password,Resource): print "Logged in as %s to server %s" % ( Username, Server ) else: print "Problems authenticating: ", con.lastErr, con.lastErrCode sys.exit(1)

Then the message and presence callbacks messageCB() and presenceCB() are defined to the connection object in con:

con.setMessageHandler(messageCB) con.setPresenceHandler(presenceCB)

After sending initial presence, informing the JSM (and anyone who might be subscribed to kassist's presence) of the assistant's availability:

con.send(jabber.Presence())

we also construct—from the Room, ConfServ, and Nick variables—and send the <presence/> element for negotiating entry to the jdev room hosted by the Conferencing component at conference.jabber.org:

roomjid = Room + '@' + ConfServ + '/' + Nick print "Joining " + Room con.send(jabber.Presence(to=roomjid))

The con.send() function will send a <presence/> element that looks like this:

SEND: <presence to='jdev@conference.jabber.org/kassist'/> We're sending available presence to the room, to negotiate entry, but what about the initial presence? Why do we send that too if there are no users who will be subscribed to the kassist JID? If no initial presence is sent, the JSM will merely store up any <message/> elements destined for kassist, as it will think the JID is offline.

The processing loop

Once everything has been set up, we simply need to have the script sit back and wait for incoming packets and handle them appropriately. For this, we simply call the process() function every 5 seconds to look for elements arriving on the XML stream:

while(1): con.process(5)

Connecting Devices to Jabber

LEGO MINDSTORMS. What a great reason to dig out that box of LEGO bricks you haven't touched in years. When I found out that LEGO was bringing out a programmable brick, the RCX,^[2], I went to my favorite toy shop and purchased the set. In addition to the RCX (shown in Figure 9-1), the MINDSTORMS set comes with an infrared (IR) port and an IR tower, which you can connect to the serial port of your PC, a battery compartment,^[3] motors, touch and light sensors, and various LEGO Technic parts.

interact with the RCX. The MINDSTORMS Robotics Invention System (RIS)™ set comes with Windows software with which you can build programs by moving blocks of logic around graphically on the screen and chaining them together. In addition, various efforts on the parts of talented individuals have come up with many different ways to program the RCX. The Unofficial Guide to LEGO® MINDSTORMS™ Robots (O'Reilly & Associates, Inc., 1999) tells you all you need to know about programing the RCX. What's important to know for this recipe is detailed in Programming the RCX.

{{Sidebar|Programming the RCXThere are two approaches to programming the RCX. One approach is to write a program on your PC, download it to the RCX, and start and stop the program using the buttons on the RCX itself.

The other approach is to control the RCX directly from a program that you write and execute on your PC, sending control signals and receiving sensor values over the IR connection.

Both approaches have their merits. How appropriate each one is boils down to one thing: connections. On the one hand, building autonomous machines that find their way around the kitchen to scare the cat and bring you a sandwich calls for the first approach, when, once you've downloaded the program to the RCX, you can dispense with any further connection with your PC because the entire logic is situated in your creation. On the other hand, if you want to build a physical extension to a larger system that, for example, has a connection to the Internet, the second approach is likely to be more fruitful, because you can essentially use the program that runs on your PC and talks to the RCX over the IR link as a conduit, a proxy of sorts, to other programs and systems that can be reached over the network. We're going to use the second approach.

The RIS software that comes as standard centers around an ActiveX control. While there are plenty of ways to talk to the RCX without using this control (the book mentioned earlier describes many of these ways), the features offered by the control—Spirit.ocx—are fine for many a project. And with Perl's Win32::OLE module, we can interact with this ActiveX control without having to resort to Visual Basic.

</code>

What We're Going to Do

Everyone knows that one of the virtues of a programmer is laziness. We're going to extend this virtue (perhaps a little too far) and enhance it with a hacker's innate ability to combine two favorite pastimes—programming and playing with LEGO—to build contrived but fun devices.

Often being a key part of a programmer's intake, coffee figures highly on the daily agenda. It's important to have a good cup of coffee to keep the brain cells firing, but it's even more important to know whether there's actually any coffee left in the pot. Going over to the coffeepot to find out is time away from the keyboard and therefore time wasted. So let's put the RCX to good use and build a device to tell us, via Jabber, whether the coffeepot has enough for another cup.

In building the device, a light sensor was connected to the RCX to "see" the level of coffee in the pot. Since the coffeepot is made of glass, light passes through it unless the coffee gets in the way, thus creating a simple binary switch:

• No (or a small amount of) light measured: there's coffee in the pot. *

Some (or a larger amount of) light: there's no coffee in the pot. We want to be able to send the availability of coffee to all interested parties in a way that their off-the-shelf Jabber clients can easily understand and display.

Figure 9-2 shows the LEGO MINDSTORMS device in action. The brick mounted on the gantry is the light sensor, which extends to the glass coffeepot; a wire runs from it to the connector on the RCX. Behind the RCX is the IR tower, which is connected to the PC.

{{Figure|title=Our device "looking" at the coffeepot|image=0596002025-jab_0902.png</code> Remembering that <presence/> elements are a simple way of broadcasting information about availability and that they contain a <status/> tag to describe the detail or context of that availability (see Section 5.4.2 for details on the <presence/> element), we have a perfect mechanism that's ready to be used. What's more, most, if not all, of the off-the-shelf Jabber client implementations will display the content of the <status/> tag in the client user's roster next to the JID to which it applies. Figure 9-3 shows how the content of the <status/> tag is displayed as a hovering "tooltip" in WinJab.

{{Figure|title=Receiving information on the coffee's status in WinJab|image=0596002025-jab_0903.png</code> Here's what we need to do:

Step 1

Set up the RCX

We need to set the RCX up, with the light sensor, so that it's close

enough to the coffeepot to take reliable and consistent light

readings. Luckily the serial cable that comes with the MINDSTORMS set

and connects to the IR tower is long enough to stretch from the

computer to within the infrared line of sight to the RCX.

Step 2

Make the correct calibrations

There are bound to be differences in ambient light, sensitivity of the

light sensor, and how strong you make your coffee. So we need a way of

calibrating the setup, so that we can find the appropriate "pivot

point" light reading value that lies between the two states of

coffee and no coffee.

Step 3

Set up a connection to Jabber

We need a connection to a Jabber server and a client account there. We

can set one up using the reguser script from Section 7.4. We also need

the script to honor presence from users who want to be informed of the

coffee state.

Step 4

Set up a sensor poll/presence push loop

Once the RCX has been set up, the sensor calibrations taken, and the

connection has been made, we need to monitor the light sensor on the

RCX at regular intervals. At each interval, we determine the coffee

state by comparing the value received from the sensor with the pivot

point determined in the calibration step and send any change in that

state as a new availability <presence/> element

containing an appropriate description in the <status/>

tag.

The Coffee Script

We're going to use Perl and the Net::Jabber libraries to build the script shown in Example 9-6. Perl allows us a comfortable way to interact with an ActiveX control, through the Win32::OLE module, so let's have a look at the coffee script as a whole, then we'll go back and look at the script in detail.

The coffee script, written in Perl

use Net::Jabber qw(Client); use Win32::OLE; use Getopt::Std; use strict;

my %opts; getopt('ls', \%opts);

use constant SERVER => "merlix.dyndns.org"; use constant PORT => 5222; use constant USERNAME => "coffee"; use constant PASSWORD => "pass"; use constant RESOURCE => "perlscript";

use constant NOCOFFEE => 0; use constant COFFEE => 1;

use constant SENSOR => defined($opts{'s'}) ? $opts{'s'} : 0; use constant GRAIN => 1;

my $current_status = -1; my @status; $status[NOCOFFEE] = 'xa/Coffeepot is empty'; $status[COFFEE] = '/Coffee is available!';

my $rcx = &setup_RCX(SENSOR);

 # Connect
 my $status = $connection->Connect( hostname => $server, port    
 => $port ); die "Cannot connect to Jabber server $server on port
 $port\n" unless $status;

 # Callbacks
 $connection->SetCallBacks( presence => \&InPresence );

 # Ident/Auth
 my @result = $connection->AuthSend( username => $user, password
 => $pass, resource => $resource ); die "Ident/Auth failed:
 $result[0] - $result[1]\n" if $result[0] ne "ok";

 # Roster
 $connection->RosterGet();

 # Initial presence dependent upon initial status
 &set_presence($connection, $initial_status);

 return $connection;

 if ($type eq "subscribe") { print "Subscribe request ($from) ...\n";
 $jabber->Send($presence->Reply(type => 'subscribed'));
 }

 if ($type eq "unsubscribe") { print "Unsubscribe request ($from)
 ...\n"; $jabber->Send($presence->Reply(type =>
 'unsubscribed'));
 }

 print <<EOT; Calibration mode. Note the sensor values and decide
 on a 'pivot' value above which 'no coffee' is signified and below
 which 'coffee' is signified.

 <STDIN>;

 while (1) { print $rcx->Poll(9, SENSOR), " "; sleep 1;
 }

 my $new_status = $val < $opts{'l'} ? COFFEE : NOCOFFEE;

 if ($new_status != $current_status) { $current_status = $new_status;
 return $current_status;
 }
 else { return undef;
 }

Examining the Coffee Script Step by Step

Now that we've seen the coffee script as a whole, let's examine it step by step to see how it works.

==== Declaring the modules, constants, and variables ==== We first declare the packages we're going to use. In addition to Net::Jabber and Win32::OLE, we're going to use Getopt::Std, which affords us a comfortable way of accepting and parsing command-line options. We also want to use the strict pragma, which should keep us from making silly coding mistakes by not allowing undeclared variables and the like.

We specify Client on the usage declaration for the Net::Jabber package to specify what should be loaded. The package is a large and comprehensive set of modules, and only some of those are relevant for what we wish to do in the script—build and work with a Jabber client connection. Other module sets are pulled in by specifying Component or Server.

use Net::Jabber qw(Client); use Win32::OLE; use Getopt::Std; use strict;

We're going to allow the command-line options -l and -s, which perform the following tasks:

No options specified (or just the -s options)

calibration mode.

When we run the script for the first time, we need to perform the

calibration and read values from the sensor to determine a midpoint

value. A number above the midpoint signifies the presence of light

and therefore the absence of coffee; below signifies the absence of

light and therefore the presence of coffee. This step is necessary

because not every environment (ambient light, sensitivity of the light

sensor, and so on) will be the same. The upper and lower values,

representing lightness and darkness, respectively, will vary across

different environments. The point is to obtain a value in between

these upper and lower values—the midpoint—with which we can compare a

light value read at any particular time. : If we don't specify any

options, the script will start up automatically in calibration mode:

C:\temp> perl coffee.pl

Figure 9-4 shows the script run in calibration mode. The values

displayed, one each second, represent the values read from the light

sensor. When the sensor was picking up lots of light, the values were

60. When the sensor was moved in front of some coffee, the values went

down to around 45. Based upon this small test, the pivot point value

was 50, somewhere in between those two values.

-l

Specify the pivot value.

Once we've determined a pivot point value, we run the script and tell

it this pivot value with the -l (light pivot):

C:\temp> perl coffee.pl -l 50

-s

Specify the sensor number.

The RCX, shown in Figure 9-1, has three connectors to which you can

attach sensors. They're the three gray 2-by-2 pieces, labeled 1, 2,

and 3, near the top of the brick. The script assumes you've attached

the light sensor to the one marked 1, which internally is 0. If you

attach it to either of the other two, you can specify the connector

using the -s (sensor) with a value of 1 (for the middle connector) or

2 (for the rightmost connector), like this:

C:\temp> perl coffee.pl -l 50 -s 2

You can specify the -s option when running in calibration or normal

modes.

{{Figure|title=Running coffee in calibration mode|image=0596002025-jab_0904.png</code> The options, summarized in Table 9-1, are defined with the Getopt::Std function:

my %opts; getopt('ls', \%opts);

Next comes a list of constants, which describe:

• The script's Jabber relationship, including the server it will connect

to and the username, password, and resource it will connect with. * The representation of the two states of coffee and no coffee, which will be used to determine the content of the <status/> tag sent along inside any <presence/> element emitted. * The identification of the connector to which the light sensor is attached and the polling granularity of the sensor (poll/presence push loop) described earlier. This item is measured in seconds.

use constant SERVER => "merlix.dyndns.org"; use constant PORT => 5222; use constant USERNAME => "coffee"; use constant PASSWORD => "pass"; use constant RESOURCE => "perlscript";

use constant NOCOFFEE => 0; use constant COFFEE => 1;

use constant SENSOR => defined($opts{'s'}) ? $opts{'s'} : 0; use constant GRAIN => 1;

The last part of the script's setup deals with the coffee state:

my $current_status = -1; my @status; $status[NOCOFFEE] = 'xa/Coffeepot is empty'; $status[COFFEE] = '/Coffee is available!';

We use a two-element array (@status) to represent the two possible coffee states. The value of each array element is a two-part string, with each part separated by a slash (/). Each of these parts will be transmitted in a <presence/> element, with the first part (which is empty in the element representing the COFFEE state) representing the presence <show/> value and the second part representing the presence <status/> value. Example 9-7 shows what a <presence/> element looks like when built up with values to represent the NOCOFFEE state.

A presence element representing the NOCOFFEE state

<presence> <show>xa</show> <status>Coffeepot is empty</status> </presence>

Most Jabber clients use different icons in the roster to represent different <show/> values. In this case, we will use xa for no coffee and a blank (which represents "online" or "available") for coffee to trigger the icon change.

Initialization and calibration

Whenever we need to talk to the RCX, some initialization is required via the ActiveX control. That's the same whether we're going to calibrate or poll for values. The setup_RCX() function takes a single argument—the identification of which connector the light sensor is connected to—and performs the initialization, which is described later in Section 9.2.3.8. The function returns a handle on the Win32::OLE object that represents the ActiveX control, which in turn represents the RCX via the IR tower:

my $rcx = &setup_RCX(SENSOR);

If the -l option is not specified, it means we're going to be running calibration. So we call the calibrate() function to do this for us. We pass the RCX handle (in $rcx) so the calibration can run properly:

# Either calibrate if no parameters given, or

As with the setup_RCX() function, calibrate() is described later.

Calibration mode will be terminated by ending the script with Ctrl-C, so the next thing we come across is the call to the function set_status(), which represents the first stage in the normal script mode; set_status() is used to determine the initial coffee status.

A value is retrieved by calling the ActiveX control's Poll() function. (Table 9-2 lists the ActiveX control's functions and properties used in this script.) We specify that we're after a sensor value (the 9 as the first argument) from the sensor attached to the connector indicated by the SENSOR constant:

# Determine initial status (will be either 0 or 1) my $s = &set_status($rcx->Poll(9, SENSOR));

The value retrieved is passed to the set_status() function, which determines whether the value is above or below the pivot value and whether the new status is different from the current one. It's going to be something along the lines of one of the values displayed when the script was run in calibration mode. If it is (and in this case, it will be, because in this first call, the value of $current_status is set to -1, which represents neither the COFFEE nor the NOCOFFEE state), that status will be returned; otherwise, undef will be returned.

==== Connecting to the Jabber server ====
At this stage, we're ready to connect to the Jabber server. The call to setup_Jabber() does this for us, returning a handle to the Jabber connection object that we store in $jabber. This handle will be used later in the script to send out <presence/> elements. The $jabber variable contains a reference to a Net::Jabber::Client object. This is the equivalent of the con variable used in the earlier Python scripts to hold the jabber.Client object and the ConnectionBean object (cb) in the earlier Java script.^[4]

my $jabber = &setup_Jabber(SERVER, PORT, USERNAME, PASSWORD, RESOURCE, $s);

In addition to passing the constants needed for the client connection to the Jabber server, we pass the initial coffee status, held in $s. We'll have a look at what the setup_Jabber() function does with this initial status a bit later when we get to the function's definition.

Sensor poll/presence push loop

Now that we've set everything up, determined the initial coffee status, and connected to the Jabber server, we're ready to start the main loop:

# Main loop: check Jabber and RCX while (1) { defined($jabber->Process(GRAIN)) or die "The connection to the Jabber server was broken\n"; my $s = &set_status($rcx->Poll(9, SENSOR)); &set_presence($jabber, $s) if defined $s; }

The while (1) loop is a bit of a giveaway. This script won't stop until you force it to by entering Ctrl-C—but that's essentially what we want. In the loop, we call the Process() method on the Jabber connection object in $jabber.

Process() is the equivalent of the Jabberpy's process() method in the Python scripts. Process() waits around for up to the number of seconds specified as the single argument (or not at all if no argument is specified) for XML to appear on the stream connection from the Jabber server. If complete fragments do appear, callbacks, defined in the connection object, are called with the elements (<iq/>, <message/>, and <presence/>) that the fragments represent. This is in the same way as, for example, callbacks are used in the Python scripts using the Jabberpy library. The setup_Jabber(), which will be discussed in the next section, is where the callback definition is made.

Net::Jabber's Process() method returns undef if the connection to the Jabber server is terminated while waiting for XML. The undef value is dealt with appropriately by ending the script.

The GRAIN constant, set to 1 second in the script's setup section, is used to specify how long to wait for any packets from the Jabber server. For the most part, we're not expecting to receive much incoming Jabber traffic—the occasional presence subscription (or unsubscription) request perhaps (see later), but other than that, the only packets traveling over the connection to the Jabber server will be availability <presence/> packets representing coffee state changes, sent from the script. This delay is normally set to 1 second. And because that's a comfortable polling interval for the light sensor, we can set that within the same loop.

Calling the ActiveX control's Poll() again with the same arguments as before ("get a sensor value from the sensor attached to the SENSORth connector"), we pass the value to the set_status() to determine the coffee state. If the state was different from last time (if $s receives a value and not undef), then we want to emit a <presence/> element to reflect that state. We achieve this by calling the set_presence() function, passing it the connection object and the state.

The setup_Jabber() function

Here we define the setup_Jabber() function, which is called to set up the connection to the Jabber server and authenticate with a predefined user:

# Set up Jabber client connection, sending intial presence sub setup_Jabber { my ($server, $port, $user, $pass, $resource, $initial_status) = @_; my $connection = new Net::Jabber::Client;

 # Connect
 my $status = $connection->Connect( hostname => $server, port    
 => $port ); die "Cannot connect to Jabber server $server on port
 $port\n" unless $status;

 # Callbacks
 $connection->SetCallBacks( presence => \&InPresence );

 # Ident/Auth
 my @result = $connection->AuthSend( username => $user, password
 => $pass, resource => $resource ); die "Ident/Auth failed:
 $result[0] - $result[1]\n" if $result[0] ne "ok";

 # Roster
 $connection->RosterGet();

 # Initial presence dependent upon initial status
 &set_presence($connection, $initial_status);

 return $connection;

First, we instantiate a new Net::Jabber::Client object. Net::Jabber distinguishes between client- and component-based connections to Jabber; the component-based equivalent of this class is Net::Jabber::Component. The Connect() method is passed arguments that specify the hostname and port of the Jabber server to connect to. It returns a 0 status if the connection could not be made.

We can register handlers for Jabber elements received over the XML stream carried by the connection we just made. Here we are interested in incoming presence subscription or unsubscription requests, as we'll see in the definition of the InPresence() function.

The single method, SetCallBacks(), does what the collective jabber.Client methods setPresenceHandler(), setMessageHandler(), and setIqHandler() do in a single call—taking a list of element types and subroutine references, in the form of a hash.

After registering the callback for <presence/> elements, it's time to authenticate, passing the username, password, and resource defined in the list of constants at the start of the script. If authentication is successful, the result of the call to the AuthSend() method is a single string with the value ok. If not, that value is replaced with an error code and the descriptive text is available in a further string. (This is why we catch the results of a call in an array, called @result.) A complete list of Jabber error codes and texts can be found in Table 5-3.

Why RosterGet()? We're not subscribing to anyone, and we're not really interested in anything but the values we're polling from our brick. So theoretically there's no reason to make a request to retrieve our roster from the server. However, because we want the script to receive and process subscription and unsubscription requests, we need to request the roster beforehand; otherwise, the JSM won't send such requests to us. See Section 8.3.3.4 in Chapter 8 for an explanation as to why.

Once we've requested the roster, so as to receive presence subscription and unsubscription requests, the job is almost done. The last thing to do in setting up the Jabber connection is to send initial availability information. The setup_Jabber() function receives the initial coffee status as the last argument in the call (in $initial_status), which it passes on to the function that sends a <presence/> element, set_presence(). Along with the initial coffee status, we also send the $connection object that represents the connection to the Jabber server that we've just established (referred to outside of this function with the $jabber variable). This is so the set_presence() function can use the connection handle to send the element down the stream.

The set_presence() function

This function is used by setup_Jabber() to send the script's (and therefore the coffee's) initial presence. It's also used within the main sensor poll/presence push loop to send further presence packets if the coffee's state changes.

sub set_presence { my ($connection, $s) = @_; my $presence = Net::Jabber::Presence->new(); my ($show, $status) = split("/", $status[$s], 2); $presence->SetPresence( show => $show, status => $status ); print $status, "\n"; $connection->Send($presence); }

On receipt of the Jabber connection object and the coffee status, which will be 0 (NOCOFFEE) or 1 (COFFEE), set_presence() constructs a new Net::Jabber::Presence object. This object represents a <presence/> element, upon which we can make method calls to hone the element as we wish. SetPresence() is one of these methods, with which we can set values for each of the <show/> and <status/> tags. We retrieve the values for each of these tags by pulling the strings from the appropriate member of the @status array, as described earlier in Section 9.2.3.1.

We print the coffee's status (remember, this function is called only when the status changes, not every time the sensor is polled) and send the newly built <presence/> element down the XML stream to the Jabber server. This is accomplished by passing the presence object as an argument to the Send() method of the connection object in $connection. This works in the same way as the send() function in Jabberpy and the send() function in JabberBeans. Everyone who has subscribed to the script user's presence, and who is available, will receive the coffee status information.

Figure 9-3 shows the status information received in the WinJab client. The string sent in the <status/> tag is shown in the tooltip that appears when the mouse hovers over the "coffee" roster item.

The InPresence() subroutine

Our presence handler, the callback subroutine InPresence(), honors requests for subscription and unsubscription to the script user's (and therefore the coffee's) presence. This callback is designed to work in the same way as the presenceCB() callback in the Python recipe described in Section 8.3.

However, while the Python Jabberpy library hands to the callbacks a jabber.Client object and the element to be handled, the Perl Net::Jabber library hands over a session ID and the element to be handled. Don't worry about the session ID here; it's related to functionality for building Jabber servers, not clients, and we can and should ignore it for the purposes of this recipe. What is important is the element to be handled, which appears as the second argument passed to the subroutine collected by the $presence variable from $_[1].

What is common between the two libraries is that the element that is passed to be handled as the subject of the callback is an instance of the class that the callback represents. In other words, a callback is used to handle <presence/> elements, and the element received is an instance of the Net::Jabber::Presence class (just as the element received by a Jabberpy presence callback is an instance of the jabber.Presence class).

# Handle presence messages sub InPresence { my $presence = $_[1]; my $from = $presence->GetFrom(); my $type = $presence->GetType();

 if ($type eq "subscribe") { print "Subscribe request ($from) ...\n";
 $jabber->Send($presence->Reply(type => 'subscribed'));
 }

 if ($type eq "unsubscribe") { print "Unsubscribe request ($from)
 ...\n"; $jabber->Send($presence->Reply(type =>
 'unsubscribed'));
 }

With an object in $presence, we can get information from the element using data retrieval methods such as those used here: GetFrom() and GetType(), which extract the values from the from and type attributes of the <presence/> element, respectively.

If the <presence/> element type represents a subscription request (type="subscribe'), we unquestioningly honor the request, by sending back an affirmative reply. The Reply() method of the presence object is one of a number of high-level functions that make it possible to turn elements around and send them back. In this case, the method replaces the value of the <presence/>'s to attribute with the value of the from attribute, and preserves its id. It also allows us to pass arguments as if we were calling the SetPresence() method described earlier. Rather than set the <show/> and <status/> tags as we did earlier in the set_presence() function, we merely set the element's type attribute to subscribed or unsubscribed, depending on the request.

So, with an incoming <presence/> element in $presence that looks like this:

<presence from='qmacro@jabber.org/office' type='subscribe' to='coffee@merlix.dyndns.org' id='21'>

calling the Reply() method would cause the element in $presence to change to this:

<presence to='qmacro@jabber.org/office' type='subscribed' id='21'>

Remember, the from attribute on elements originating from the client is set by the server, not by the client. The script doesn't ask for a subscription to the user's presence in return. The script isn't interested in whether the people who have subscribed to its presence are available—its purpose is to let people know whether there's any coffee left in the pot.

The setup_RCX() function

This function is called once every time the script is started and is required to initialize the RCX:

sub setup_RCX { my $sensor = shift; my $rcx = Win32::OLE->new('SPIRIT.SpiritCtrl.1'); $Win32::OLE::Warn = 0; $rcx->{ComPortNo} = 1; $rcx->{InitComm}; $rcx->SetSensorType($sensor, 3); $rcx->SetSensorMode($sensor, 2); return $rcx; }

A Win32::OLE object representing the RCX's ActiveX control Spirit is instantiated. A Win32::OLE function is used to suppress warnings, and the RCX is initialized by setting the COM port to COM1 for serial communications. The sensor type and mode are set for the light sensor attached to the connector identified by the value passed into the $sensor variable. Table 9-2 shows us that sensor type 3 represents Light, and sensor mode 2 specifies a Transitional measurement mode, the upshot of which is that the values returned on a poll are all within a certain restricted range, which makes it easier to decide whether there's any coffee in the pot.

We return the Win32::OLE RCX object to be used elsewhere in the script for calibration and polling.

The calibrate() function

The calibrate() function is called if the script is started without the -l option. This function simply prints a message, waits for the user to press Enter, and then goes into a a gentle loop, emitting whatever value was polled from the light sensor so the user can determine the pivot point:

sub calibrate { my $rcx = shift;

 print <<EOT; Calibration mode. Note the sensor values and decide
 on a 'pivot' value above which 'no coffee' is signified and below
 which 'coffee' is signified.

 <STDIN>;

 while (1) { print $rcx->Poll(9, SENSOR), " "; sleep 1;
 }

The output produced from this function can be seen in Figure 9-4.

The set_status() function

The set_status() function receives the latest light value as polled from the sensor and compares it with the pivot value. If the status defined in $new_status is different from the current status (in $current_status), then the current status is updated and returned; otherwise, undef is returned:

sub set_status { my $val = shift;

 my $new_status = $val < $opts{'l'} ? COFFEE : NOCOFFEE;

 if ($new_status != $current_status) { $current_status = $new_status;
 return $current_status;
 }
 else { return undef;
 }

If this function returns a status value, a new <presence/> element is generated and emitted by the script. Otherwise, there's no change ("the coffee's still there," or "there's still no coffee!") and nothing happens.

An RSS News Agent

While the Jabber clients available off the shelf are orientated toward receiving (and sending) messages from other people, the possibilities don't stop there, as is clear from the recipes we've seen already. In this recipe, we're going to build a Jabber component that retrieves news items from various sources on the Web and sends them on to Jabber users who have expressed an interest in receiving them. We're going to use the Web for our news sources, but they could just as easily be sources within a corporate intranet. The key thing is that the sources are available in a readily parseable format.

RSS (RDF^[5] Site Summary or, alternatively, Really Simple Syndication) is an XML format used for describing the content of a web site, where that site typically contains news items, diary entries, event information, or generally anything that grows, item by item, over time. A classic application of RSS is to describe a news site such as JabberCentral (http://www.jabbercentral.org). JabberCentral's main page (see Figure 9-5) consists of a number of news items—in the "Recent News" section—about Jabber and its developer community. These items appear in reverse chronological order, and each one is succinct, sharing a common set of properties:

Title

Each item has a title ("JabberCon Update 11:45am - Aug 20").

Short description

Each item contains a short piece of text describing the content and

context of the news story ("JabberCon Update - Monday Morning").

Link to main story

The short description should be enough to help the reader decide if he

wants to read the whole item. If he does, there's a link ("Read More")

to the news item itself.

{{Figure|title=JabberCentral's main page|image=0596002025-jab_0905.png</code> It is this collection of item-level properties that are summarized in an RSS file. The formality of the XML structure makes it a straightforward matter for:

• Automating the retrieval of story summaries for inclusion in other

sites (syndication) * Combining these items with items from other similar sources (aggregation) * Checking to see whether there is any new content (new items) since the last visit Example 9-8 shows what the RSS XML for JabberCentral's news items shown in Figure 9-5 looks like.

RSS source for JabberCentral

<?xml version="1.0" encoding="ISO-8859-1"?>

<!DOCTYPE rss PUBLIC "-//Netscape Communications//DTD RSS 0.91//EN" "http://my.netscape.com/publish/formats/rss-0.91.dtd">

<rss version="0.91">

 <channel>

   <title>JabberCentral</title>

   <description> JabberCentral is the premiere Jabber end-user
   news and support site. Many Jabber developers are actively involved
   at JabberCentral to provide fresh and authoritative information for
   users. </description>

   <language>en-us</language>
   <link>http://www.jabbercentral.com/</link>
   <copyright>Copyright 2001, Aspect Networks</copyright>

   <image>
   <url>http://jabbercentral.com/images/jc_button.gif</url>
   <title>JabberCentral</title>
   <link>http://www.jabbercentral.com/</link>
   </image>

   <item> <title>JabberCon Update 11:45am - Aug
   20</title>
   <link>http://www.jabbercentral.com/news/view.php?news_id=
   998329970</link> <description>JabberCon Update - Monday
   Morning</description> </item>

   <item> <title>Jabcast Promises Secure Jabber
   Solutions</title>
   <link>http://www.jabbercentral.com/news/view.php?news_id=
   998061331</link> <description> Jabcast announces their
   intention to release security plugins with their line of products
   and services. </description> </item>

   ... (more items) ...

 </channel>

The structure is very straightforward. Each RSS file describes a channel, which is defined as follows:

Channel information

The channel header information includes the channel's title

(<title/>), short description

(<description/>), main URL (<link/>),

and so on. The channel in this case is JabberCentral.

Channel image

Often RSS information is rendered into HTML to provide a concise

"current index" summary of the channel it describes. An image can be

used in that summary rendering, and its definition is held in the

<image/> section of the file.

Channel items

The bulk of the RSS file content is made up of the individual

<item/> sections, each of which reflects an item on the

site that the channel represents. We can see in Example 9-8 that the

first <item/> tag:

<item> <title>JabberCon Update 11:45am - Aug 20</title> <link>http://www.jabbercentral.com/news/view.php?news_id=998329970 </link> <description>JabberCon Update - Monday Morning</description> </item>

describes the most recent news item shown on JabberCentral's main

page—"JabberCon Update 11:45am - Aug 20." Each of the news item

properties are contained within that <item/> tag: the

title (<title/>), short description

(<description/>), and link to main story

(<link/>).

Channel interactive feature

There is a possibility for each channel to describe an interactive

feature on the site it represents; often this is a search engine

fronted by a text input field and Submit button. The interactive

feature section of an RSS file is used to describe how that mechanism

is to work (the name of the input field and the Submit button and the

URL to invoke when the button is clicked, for example). This is so

HTML renderings of the site can include the feature otherwise

available only on the original site.

{{Note|This interactive feature definition is not shown in the RSS example here.

</code> RSS information lends itself very well to various methods of viewing. There are custom "headline viewer" clients available—focused applications that allow you to select from a vast array of RSS sources and have links to items displayed on your desktop (so, yes, the personal newspaper—of sorts—is here!). There are also possibilities for having RSS items scroll by on your desktop control bar.

And then there's Jabber. As described in Section 5.4.1, the Jabber <message/> element can represent something that looks suspiciously like an RSS item. The message type "headline" defines a message that carries news headline information. In this case, the <message/> element itself is usually embellished with an extension, qualified by the jabber:x:oob namespace (described in Section 6.3.8). Example 9-9 shows what the element would look like if the first news item from the JabberCentral site were carried in a headline message.

A headline message carrying a JabberCentral news item

<message type='headline' to='dj@qmacro.dyndns.org'> <subject>JabberCon Update 11:45am - Aug 20</subject> <body>JabberCon Update - Monday Morning</body> <x xmlns='jabber:x:oob'> <url>http://www.jabbercentral.com/news/view.php?news_id=998329970& lt;/url> <desc>JabberCon Update - Monday Morning</desc> </x> </message>

The jabber:x:oob namespace carries the crucial parts of the RSS item. Clients, such as WinJab and Jarl, can understand this extension and display the content in a clickable list of headlines, each representing a single RSS item, similar to the headline viewer clients mentioned earlier.

Of course, we could send RSS items to clients in nonheadline type messages:

<message type='headline' to='dj@qmacro.dyndns.org'> <subject>JabberCon Update 11:45am - Aug 20</subject> <body> JabberCon Update - Monday Morning http://www.jabbercentral.com/news/view.php?news_id=998329970 </body> </x> </message>

where the complete item information is transmitted in a combination of the <subject/> and <body/> tags. This works, but the clients can only display the message, as with any other message. However, if we send formalized metadata, the value of the message content increases enormously. (Figure 9-8 shows Jarl displaying RSS-sourced news headlines.)

Distributing RSS-sourced headlines over Jabber to standard Jabber clients is a great combination of off-the-shelf technologies. In fact, we'll see in the next section that it's not just standard Jabber clients that fit the bill; we'll write a Jabber-based headline viewer to show that not all Jabber clients are, nor should they be, made equal.

Writing the News Agent

We're going to write an RSS news agent, which we'll simply call newsagent. The newsagent is a mechanism that checks predefined sources for new RSS items and sends (or pushes) them to people who are interested in receiving them. For the sake of simplicity, we'll define the list of RSS sources in newsagent itself. See Section 9.3.4 later in this chapter for details on how to further develop this script.

The newsagent script as a component

Until now, the examples we've used, such as cvsmsg, HostAlive, and keyassist (shown in Chapter 8), have all existed as Jabber clients. That is, they've performed a service while connected to the Jabber server via the JSM. There's nothing wrong with this. Indeed, it's more than just fine to build Jabber-based mechanisms using a Jabber client stub connection; that way, your script, through its identity—the user JID—can avail itself of all the IM-related functions that the JSM offers—presence, storage and forwarding of messages, and so on. Perhaps even more interesting is that the mechanism needs only an account, a username, and a password on a Jabber server to be part of the big connected picture.

However, we know from Chapter 4 that there are other entities that connect to Jabber to provide services. These entities are called components. You can look at components as philosophically less "transient" than their client-connected brethren and also closer to the Jabber server in terms of function and connection.

We know from Section 4.1.1 that there are various ways to connect a component: library load, STDIO, and TCP sockets. The first two dictate that the component will be located on the same host as the jabberd backbone to which it connects, although a Jabber server could consist of

a collection of jabberds running on separate hosts.    The TCP

sockets connection type uses a socket connection between the component and the jabberd backbone, over which streamed XML documents are exchanged (in the same way they are exchanged in a client connection). It allows us to run components on any host and connect them to a Jabber server running on another host if we wish. This approach is the most desirable because of the connection flexibility. But it's not just the flexibility that matters: because the component is abstracted away from the Jabber server core libraries, it's up to us to decide how the component should be written. All the component has to do to get the Jabber server to cooperate is to establish the socket connection as described in the component instance configuration, perform an authenticating handshake, and correctly exchange XML stream headers.

Let's review how a TCP socket-based component connects. We'll base the review on what we're actually going to have to do to get newsagent up and running.

First, we have to tell the Jabber server that it is to expect an incoming socket connection attempt, which it is to accept. We do this by defining a component instance definition (or "description"—see Section 4.2.1) for our component. We include this definition in the main Jabber server configuration file, usually called jabber.xml. Example 9-10 shows a component instance definition for the RSS news agent, known as rss.qmacro.dyndns.org.

A component instance definition for the RSS news agent

<service id='rss.qmacro.dyndns.org'> <accept> <ip>localhost</ip> <port>5999</port> <secret>secret</secret> </accept> </service>

The name of the host on which the main Jabber server is running is qmacro.dyndns.org; it just so happens that the plan is to run the RSS news agent component on the same host. We give it a unique name (rss.qmacro.dyndns.org) to enable the jabberd backbone, or hub, to distinguish it from other components and to be able to route elements to it.

An alternate way of writing the component instance definition is shown in Example 9-11. The difference is simply in the way we specify the name. In Example 9-10, we specified an id in the <service/> tag with the value rss.qmacro.dyndns.org. In the absence of any <host/> tag specification in the definition, this id value is used by the jabberd routing logic as the identification for the component when determining where elements addressed with that destination should be sent. In Example 9-11, we have an explicit <host/> specification that will be used instead; we simply identify the service with an id attribute value of rss. In this latter case, it doesn't really matter from an addressability point of view what we specify as the value for the id attribute.

An alternative instance definition for the RSS news agent

<service id='rss'> <host>rss.qmacro.dyndns.org</host> <accept> <ip>localhost</ip> <port>5999</port> <secret>secret</secret> </accept> </service>

The instance definition contains all the information the Jabber server needs. We can tell from the <accept/> tag that this definition describes a TCP sockets connection. The socket connection detail is held in the <ip/> and <port/> tags. In this case, as we're going to run the RSS News Agent component on the same host as the Jabber server itself, we might as well kill two related birds—performance and security—with one stone by specifying in the <ip/> tag:^[6]

Performance

Connecting over the loopback device, as opposed to a real network

interface, will give us a slight performance boost.

Security

Accepting only on the loopback device is a simple security measure

that leaves one less port open to the world. The

<secret/> tag holds the secret that the connecting

component must present in the authentication handshake. How the secret

is specified is described later on in this section.

Now let's look at the component's view of things. It will need to establish a socket connection to 127.0.0.1:5999. Once that connection has been established, jabberd will be expecting it to announce itself by sending its XML document stream header. Example 9-12 shows a typical stream header that the component will need to send.

The RSS component's stream header

SEND: <?xml version='1.0'?> <stream:stream xmlns='jabber:component:accept' xmlns:stream='http://etherx.jabber.org/streams' to='localhost'>

This matches the description of a Jabber XML stream header (also known as a stream "root" as it's the root tag of the XML document) from Section 5.3. The namespace that is specified as the one qualifying the content of the stream is jabber:component:accept. This namespace matches the component connection method (TCP sockets) and the significant tag name in the component instance definition (). Likewise, the namespace jabber:component:exec matches the STDIO component connection method and the significant tag name in its component instance definition format: ()—see Section 4.1.3.3. The value specified in the to attribute matches the hostname specified in the configuration's <ip/> tag.

After receiving a valid stream header, jabberd responds with a similar root to head up its own XML document stream going in the opposite direction (from server to component). A typical response to the header (Example 9-12) received from the server by the component is shown in Example 9-13.

The server's stream header reply

RECV: <?xml version='1.0'?> <stream:stream xmlns:stream='http://etherx.jabber.org/streams' id='3B8E3540' xmlns='jabber:component:accept' from='rss'>

The stream header sent in response shows that the server is confirming the component instance's identification as rss. This reflects whatever was specified in the <service/> tag's id attribute of the component instance definition. Here, the value of the id attribute was rss as in Example 9-11.

It also contains an ID for the component instance itself (id="3B8E3540'). This ID is a random string shared between both connecting parties; the value is used in the next stage of the connection attempt—the authenticating handshake.

The digest authentication method for clients connecting to the JSM is described in Section 7.3.1.2. This method uses a similar shared random string. On receipt of the server's stream header, the component takes the ID and prepends it onto the secret that it must authenticate itself with. It then creates a NIST SHA-1 message digest (in a hexadecimal format) of that value:

SHA1_HEX(ID+SECRET) After the digest is created, it is sent in a <handshake/> element as the first XML fragment following the root:

SEND: <handshake id="1">14d437033d7735f893d509c002194be1c69dc500</handshake>

On receipt of this authentication request, jabberd combines the ID value with the value from the <secret/> tag in the component instance definition and performs the same digest algorithm. If the digests match, the component is deemed to have authenticated itself correctly, and it is then sent back an empty <handshake/> tag in confirmation:

<handshake/> The component may commence sending (and being sent) elements.

If the component sends an invalid handshake value—the secret may be wrong or the digest may not have been calculated correctly—the connection is closed: jabberd sends a stream error, ending the conversation:

RECV: <stream:error>Invalid handshake</stream:error>

==== Working out who gets what newsfeeds ==== Definitions of the RSS sources are held within the newsagent itself, but there's no reference to who might want to receive new items from which sources. We need a way for the component to accept requests, from users, that say things like:

"I'd like to have pointers to new items from Slashdot sent to me, please."

"I'd also like pointers to new items on Jon Udell's site, please."

"Whoa, information overflow! Stop all my feeds!"

as the Jabber User Directory (JUD), and the transports to other IM systems such as Yahoo! and ICQ. This theme is known as registration. We've seen this before in the form of user registration, described in Section 6.5.2. This is the process of creating a new account with the JSM. Registration with a service such as the JUD or an IM transport, however, follows a similar process, and both types of registration have one thing in common: the jabber:iq:register namespace.

The jabber:iq:register namespace is used to qualify the exchange of information during a registration process. The registration process to create a new user account with the JSM uses the jabber:iq:register namespace to qualify registration data exchanged. The registration process with the JSM to modify the account details (name, email address, and so on) also uses jabber:iq:register to qualify the account amendment data exchanged. Both types of registration requests are addressed to the JSM. The difference, which allows the JSM to distinguish between what is being requested, is that no session is active on the stream between client and server in the new user registration process, whereas in the account amendment process, a session is active. This is also mentioned in Section 7.2.2.5Chapter 7.

The jabber:iq:register namespace is described in Section 6.2.11 in Chapter 6. It shows us how a typical conversation between requester and responder takes place:

1. The client sends an IQ-get: "How do I register?" The component
2. sends an IQ-result: "Here's how:

follow these instructions to fill in these fields."

1. The client then sends an IQ-set with values in the fields: "OK,
2. here's my registration request." To which the component responds,
3. with another IQ-result: "Looks fine. Your registration details have
4. been stored."

It's clear that this sort of model will lend itself well to the process of allowing users to make requests to receive pointers to new items from RSS sources chosen from a list. Example 9-14 shows this conversational model in Jabber XML. There are many fields that can be used in a registration request; the description in Section 6.2.11 in Chapter 6 includes a few of these—<name/>, <first/>, <last/>, and <email/>—but there are more. We'll take the <text/> field to accept the name of an RSS source when a user attempts to register his interest to receive pointers to new items from that source. The conversational model is shown from the component's perspective.

A registration conversation for RSS sources "How do I register?"

RECV: <iq type='get' id='JCOM_3' to='rss.qmacro.dyndns.org' from='dj@qmacro.dyndns.org/basement'> <query xmlns='jabber:iq:register'/> </iq>

"Here's how:"

SEND: <iq id='JCOM_3' type='result' to='dj@qmacro.dyndns.org/basement' from='rss.qmacro.dyndns.org'> <query xmlns='jabber:iq:register'> <instructions> Choose an RSS source from: Slashdot, Jon Udell[, ...] </instructions> <text/> </query> </iq>

"OK, here's my registration request:"

RECV: <iq type='set' id='JCOM_5' to='rss.qmacro.dyndns.org' from='dj@qmacro.dyndns.org/basement'> <query xmlns='jabber:iq:register'> <text>Slashdot</text> </query> </iq>

"Looks fine. Your registration details have been stored."

SEND: <iq id='JCOM_5' type='result' to='dj@qmacro.dyndns.org/basement' from='rss.qmacro.dyndns.org'> <query xmlns='jabber:iq:register'> <text>Slashdot</text> </query> </iq>

After some time passes...

"Whoa, information overflow! Stop all my feeds!"

RECV: <iq id='JCOM_11' to='rss.qmacro.dyndns.org' type='set' from='dj@qmacro.dyndns.org/basement'> <query xmlns='jabber:iq:register'> <remove/> </query> </iq>

"OK, you've been removed. All feeds stopped."

SEND: <iq id='JCOM_11' to='dj@qmacro.dyndns.org/basement' type='result' from='rss.qmacro.dyndns.org'> <query xmlns='jabber:iq:register'> <remove/> </query> </iq>

A lightweight persistent storage system is used for the user/source registrations—DataBase Manager (DBM)—to keep the script fairly simple.

The bigger question here is: how will the user know he can register to a particular RSS feed? Or more importantly: how can he determine if the RSS News Agent system exists? Most clients, having connected to the server and established a session with the JSM, make a request for a list of agents (old terminology) or services (new terminology) available from the Jabber server with the following IQ-get method:

SEND: <iq id="wjAgents" to="qmacro.dyndns.org" type="get"> <query xmlns="jabber:iq:agents"/> </iq>

The response to the request looks like this:

RECV: <iq id='wjAgents' to='dj@qmacro.dyndns.org/basement' type='result' from='qmacro.dyndns.org'> <query xmlns='jabber:iq:agents'> <agent jid='conf.qmacro.dyndns.org'> <name>Public Chatrooms</name> <service>public</service> <groupchat/> </agent> <agent jid='users.jabber.org'> <name>Jabber User Directory</name> <service>jud</service> <search/> <register/> </agent> </query> </iq>

which reflects the contents of the <browse/> section in the JSM configuration as shown in Example 9-15.

The JSM configuration's browse section

<browse> <conference type="public" jid="conf.qmacro.dyndns.org" name="Public Chatrooms"/> <service type="jud" jid="users.jabber.org" name="Jabber User Directory"> <ns>jabber:iq:search</ns> <ns>jabber:iq:register</ns> </service> </browse>

If we add a stanza that describes the component for the RSS News Agent to the <browse/> section of the JSM configuration:

<service type="rss" jid="rss.qmacro.dyndns.org" name="RSS News Agent"> <ns>jabber:iq:register</ns> </service>

we get an extra section in the jabber:iq:agents response from the server:

<agent jid='rss.qmacro.dyndns.org'> <name>RSS News Agent</name> <service>rss</service> <register/> </agent>

The client-side effect of the agents response is exactly what we're looking for. Figure 9-6 shows WinJab's Agents menu displaying a summary of what it received in response to its jabber:iq:agents query.

{{Figure|title=WinJab's Agents menu|image=0596002025-jab_0906.png</code> We can see that the stanza for the RSS news agent was present in the <browse/> section and the component is faithfully displayed in the agent list, along with Public Chatrooms and Jabber User Directory. In the main window of the screenshot we can see the Supported Namespaces list; it contains the namespace that we specified in the stanza. By specifying:

<ns>jabber:iq:register</ns> we're effectively telling the client that the component will support a registration conversation.

But that's not all—we've advertised the RSS news agent in the <browse/> section of the configuration for the JSM on the Jabber server running on qmacro.dyndns.org. That's why we got the information about the RSS news agent when we connected as user dj to qmacro.dyndns.org (see the window's titlebar in Figure 9-6). You may have noticed something odd about the definition of the other two agents, or services, in the <browse/> section earlier or in the corresponding jabber:iq:agents IQ response. Let's take a look at this response again, this time with the extra detail about the component:

RECV: <iq id='wjAgents' to='dj@qmacro.dyndns.org/basement' type='result' from='qmacro.dyndns.org'> <query xmlns='jabber:iq:agents'> <agent jid='rss.qmacro.dyndns.org'> <name>RSS News Agent</name> <service>rss</service> <register/> </agent> <agent jid='conf.qmacro.dyndns.org'> <name>Public Chatrooms</name> <service>public</service> <groupchat/> </agent> <agent jid='users.jabber.org'> <name>Jabber User Directory</name> <service>jud</service> <search/> <register/> </agent> </query> </iq>

While the jid attribute values for the RSS news agent and Public Chatroom agents show that they are components connected to the Jabber server (i.e., they both have JIDs in the qmacro.dyndns.org "space," and so are connected to the Jabber server running at qmacro.dyndns.org), the jid attribute for the Jabber User Directory points to a name in the jabber.org "space"! This is a side effect of the power and foresight of Jabber's architectural design. If we connect a component—whether it's one we've built ourselves or one we've downloaded from http://download.jabber.org—we can give it an internal or external identity when we describe it in the jabber.xml configuration.

Example 9-8 and Example 9-9 show two examples of an instance definition for the RSS news agent component. Both specify potentially external identities. If the hostname rss.qmacro.dyndns.org is a valid and resolvable hostname, the component can be reached from anywhere, not just from within the Jabber server to which it is connected. If the hostname wasn't resolvable by the outside world, by having a simple name such as rss, it could be reached only from the Jabber server to which it was connected.

So let's say rss.qmacro.dyndns.orgis a valid and resolvable hostname. If your client is connected to a Jabber server running on yourserver.org, this is what would happen if you were to send a registration request (an <iq/> element with a query qualified by the jabber:iq:register namespace) addressed to rss.qmacro.dyndns.org:

Packet reaches JSM on yourserver.org.

You send the IQ from your client, which is connected to your Jabber

server's JSM. This is where the packet first arrives.

Internal routing tables consulted.

This is how yourserver.org's jabberd looks in its list of

internally registered destinations and doesn't find

rss.qmacro.dyndns.org in there.

Name resolved and routing established.

yourserver.org's dnsrv (Hostname Resolution) service

is used to resolve the rss.qmacro.dyndns.org's address. Then,

according to dnsrv's instance configuration (specifically the

<resend>s2s</resend> part—see Section 4.9), the IQ is

routed on to the s2s (Server to Server) component.

Server to server connection established.

yourserver.org establishes a connection to

qmacro.dyndns.org via s2s and sends the IQ across

the connection.

Packet arrives at the RSS News Agent component on qmacro.dyndns.org.

jabberd on qmacro.dyndns.org routes the packet correctly to

rss.qmacro.dyndns.org. So, what do we learn from this? As

exemplified by the reference to the JUD running at

users.jabber.org that comes predefined in the standard

jabber.xml with the 1.4.1 version of the Jabber server, you can

specify references to services, components, on other Jabber

servers. If you take this RSS News Agent script and run it against

your own Jabber server, there's no reason why you can't share its

services with your friends who run their own Jabber servers.

The key is not the reference in the <browse/> section; it is the resolvability of component names as hostnames and the ability of Jabber servers to route packets to each other. The stanza in <browse/> just makes it easier for the clients to automatically know about and be able to interact with services in general. Even if a service offered by a public component that wasn't described in the result of a jabber:iq:agents query, it wouldn't stop you from reaching it. The agent browser is another client, Gabber (shown in Figure 9-7), which is a GTK-based Jabber client that allows you to specify a Jabber server name, in the Server to Browse field, so that you can direct the jabber:iq:agents queries to whatever server you want.

{{Figure|title=Viewing agents on different Jabber servers with Gabber|image=0596002025-jab_0907.png</code> A good example of the distinction between the definition of a component within a <browse/> section and that component's reachability is the version query shown in Example 9-16. Regardless of whether the conference component at gnu.mine.nu was listed in the <browse/> section of qmacro.dyndns.org's JSM, the user dj was able to make a version query by specifying the component's address (a valid and resolvable hostname) in the IQ-get's to attribute.

A Conferencing component responds to a version query

SEND: <iq type='get' to='conf.gnu.mine.nu'> <query xmlns='jabber:iq:version'/> </iq>

RECV: <iq type='result' to='dj@qmacro.dyndns.org/study' from='conf.gnu.mine.nu'> <query xmlns='jabber:iq:version'> <name>conference</name> <version>0.4</version> <os>Linux 2.2.13</os> </query> </iq>

==== Polling the RSS sources ==== Next, we need some way of "interrupting" the process of checking for incoming elements and dispatching them to the callbacks, while we retrieve the RSS data and check for new items. Since we're writing this component in Perl, we could use the alarm() feature to set an alarm and have a subroutine invoked, to poll the RSS sources, when the alarm goes off. However, this recipe uses the Jabber::Connection library, which negates the needs for an external alarm. Instead, we need to take the following steps each time we want to poll the RSS sources:

1. Try to retrieve the source from the URL. Attempt to parse the source's
2. XML. Go through the items, until we come across one we've seen before;
3. the ones we go through until then are deemed to be new. (We need a
4. special case the first time around, so that we don't flood everyone
5. with every item of a source the first time it is retrieved.) For new
6. items, look in the registration database for the users that have
7. registered for that source, construct a headline message like the one
8. shown in Example 9-7, and send it to those users. Remember the first
9. of the new items, so that we don't go beyond it next time.

=== Other Differences Between Client and Component Programming === There are many differences between programming a component and programming a client. We're already aware of many of the major ones, described earlier in Section 9.3.1.1. There are, however, also more subtle differences that we need to bear in mind.

Components, unlike clients, do not connect to the JSM. They connect as a peer of the JSM. This means not only that they cannot partake of IM features made available by JSM's modules (see Section 4.4.4 for a list of these modules) but also that they must do more for themselves. This isn't as bad as it seems. Take store and forward, for example, a feature provided by the JSM's mod_offline module. While a message sent to a component won't be stored and forwarded if that component is not connected, a message sent from a component to a client will get stored and forwarded if the client is offline, because the message will be routed to the JSM (as specified by the [hostname] in the address), which can decide what action to take. Messages can be passed directly to the client if the user is online or can be stored and forwarded later when they're back online.

When constructing an element as a client, we should not specify a from attribute before it is sent; this is added by the JSM as it arrives to prevent JID spoofing. If a component does not connect through the JSM, no "from-stamping" takes place; the component itself must stamp the element with a from attribute.

The addressing of a component is also slightly different. Client addresses reflect the fact that they're connected to the JSM, always having the form of (with the resource being optional):

[user]@[hostname]/[resource] While the basic address form of a component is simply:

[hostname] This doesn't mean to say that the address of a component cannot have a user or a resource part. It's just that all elements addressed to:

anything@[hostname]/anything will be routed by jabberd to the component. This means the component can play multiple roles and have many personalities. We'll see an example of this in the script, where we construct an "artificial" [user]@[hostname] address for the from attribute of a <message/> element, to convey information.

The component will respond to IQ queries in the jabber:iq:register namespace. It is customary, although by no means mandatory, for components to respond to queries in a set of common IQ namespaces. We see that both the JUD and Conferencing components, for example, respond to IQ queries in the jabber:iq:time and jabber:iq:version namespaces. Example 9-16 shows a typical version query on a Conferencing component. This responsiveness is simply to provide a basic level of administrative information. We want the component to conform to the customs, so we'll make sure it also responds to queries in these namespaces.

=== The RSS News Agent Script === The RSS news agent script (newsagent) is written in Perl and is shown here in Example 9-17.

The RSS news agent script, written in Perl

my $NAME = 'RSS News Agent'; my $ID = 'rss.qmacro.dyndns.org'; my $VERSION = '0.1'; my $reg_file = 'registrations'; my %reg;

my %cache;

my %sources = ( 'jonudell' => 'http://udell.roninhouse.com/udell.rdf', 'slashdot' => 'http://slashdot.org/slashdot.rdf',

 # etc ...

 # Reg query
 if ($node->attr('type') eq IQ_GET) { $node = toFrom($node);
 $node->attr('type', IQ_RESULT); my $instructions = "Choose an RSS
 source from: ".join(", ", keys %sources);
 $query->insertTag('instructions')->data($instructions);
 $query->insertTag('text'); $c->send($node);
 }

 # Reg request
 if ($node->attr('type') eq IQ_SET) {

   # Strip JID to user@host
   my $jid = stripJID($node->attr('from'));

   $node = toFrom($node); my $source;

   # Could be an unregister
   if ($query->getTag('remove')) { delete $reg{$jid};
   $node->attr('type', IQ_RESULT);
   }

   # Otherwise it's a registration for a source
   elsif ($source = $query->getTag('text')->data and
   exists($sources{$source})) { my $element = $reg{$jid};
   $element->{$source} = 1; $reg{$jid} = $element;
   $node->attr('type', IQ_RESULT);
   }

   else { $node->attr('type', IQ_ERROR); my $error =
   $node->insertTag('error'); $error->attr('code', '405');
   $error->data('Not Allowed');
   }

   $c->send($node);
 }
 return r_HANDLED;

 # Create NodeFactory
 my $nf = new Jabber::NodeFactory;

 # Go through each of the RSS sources
 foreach my $source (keys %sources) {

   # Retrieve attempt
   my $data = get($sources{$source});

   # Didn't get it? Next one
   unless (defined($data)) { debug("Cannot retrieve $source"); next;
   }

   # Parse the RSS
   my $rss = XML::RSS->new(); eval { $rss->parse($data) };

   if ($@) { debug("Problems parsing $source"); next;
   }

   my @items = @{$rss->{items;

   # Check new items
   debug("$source: looking for new items"); foreach my $item (@items) {

     # Stop checking if we get to items already seen
     last if exists $cache{$source} and $cache{$source} eq
     $item->{link};

     debug("$source: new item $item->{title}");

     # Create a headline message
     my $msg = $nf->newNode('message');

     $msg->attr('type', 'headline'); $msg->attr('from', join('@',
     $source, $ID));
     $msg->insertTag('subject')->data($item->{title});
     $msg->insertTag('body')->data($item->{description});

     my $xoob = $msg->insertTag('x', NS_XOOB);
     $xoob->insertTag('url')->data($item->{link});
     $xoob->insertTag('desc')->data($item->{description});

     # Deliver to all that want it
     foreach my $jid (keys %reg) {

       my $registration = $reg{$jid};

       if (exists($registration->{$source})) { $msg->attr('to',
       $jid); debug("Sending to $jid"); $c->send($msg);
       }
     }

     # Prevent all items counted as new the first time around
     last unless exists($cache{$source});
   }

   # Remember the latest new item
   $cache{$source} = $items[0]->{link};
 }

}

sub cleanup { debug("Cleaning up"); untie %reg; $c->disconnect; exit; }

sub toFrom { my $node = shift; my $to = $node->attr('to'); $node->attr('to', $node->attr('from')); $node->attr('from', $to); return $node; }

sub stripJID { my $JID = shift; $JID =~ s|/.*$||; return $JID; }

sub debug { print STDERR "debug: ", @_, "\n"; }

</code>

=== Reviewing the RSS News Agent Script Step by Step === Now that we know what the newsagent script looks like in its entirety, let's review the script piece by piece.

==== Module declarations and variable definitions ==== We start out by using the Jabber::Connection library, which is defined as follows:

use strict; use Jabber::Connection; use Jabber::NodeFactory; use Jabber::NS qw(:all); use MLDBM 'DB_File'; use LWP::Simple; use XML::RSS;

The Jabber::Connection library consists of the following three modules:

Jabber::Connection

This module is used to manage the connection to the server and parses

and dispatches incoming elements.

Jabber::NodeFactory

This module allows us to manipulate elements, which are generically

called nodes.

Jabber::NS

The last module provides us with a list of constants that reflect

namespaces and other common strings used in Jabber server, client, and

component programming. Next we need a way of storing the registration

information between invocations of the component script, and for that

we'll use the Multi-Level Database Manager module, MLDBM.

MLDBM is a useful wrapper that can be placed around the

DB_File module. DB_File provides access to Berlekey

Database (Berkeley DB, at http://www.sleepycat.com) facilities using

the tie() function. While you can't store references (i.e.,

complex data structures) via DB_File, you can with the

MLDBM wrapper.

We will use the LWP::Simple module to grab the RSS sources by URL and the XML::RSS module to parse those sources once retrieved:

my $NAME = 'RSS News Agent'; my $ID = 'rss.qmacro.dyndns.org'; my $VERSION = '0.1'; my $reg_file = 'registrations'; my %reg;

my %cache;

my %sources = ( 'jonudell' => 'http://udell.roninhouse.com/udell.rdf', 'slashdot' => 'http://slashdot.org/slashdot.rdf',

 # etc ...

We start by declaring a few variables. We will see later in the script that $NAME, $ID, and $VERSION will be used to reflect information in response to IQ queries. The variable $reg_file defines the name of the DB file to which we will be tying the registration hash, %reg. %cache is the RSS item cache, which holds items we've already seen so we know when we've come to the end of the new items in a particular source.

We define the RSS sources in %sources. These can be defined differently, perhaps outside of the script, but this gives you a general idea of how these should look. There are a couple of examples here; add your own favorite channels to taste.

To make persistent any data we store in the %reg hash, we can use the magic of the tie() function:

tie (%reg, 'MLDBM', $reg_file) or die "Cannot tie to $reg_file: $!\n";

It works by binding (tie'ing) the operations on the hash (add, delete, and so on) to Berlekey DB operations, using the MLDBM module to stringify (and reconstruct) complex data structures so they can be stored and retrieved.

==== Connecting to the Jabber server ==== Now we're ready to connect to the Jabber server as a component. Despite what's involved (described in Section 9.3.1.1) it's very easy to make the connection to the Jabber server using a library such as Jabber::Connection:

my $c = new Jabber::Connection( server => 'localhost:5999', localname => $ID, ns => 'jabber:component:accept', );

We construct a Jabber::Connection object, specifying the details of the connection we wish to make. The server argument is used to specify the hostname, and optionally the port, of the Jabber server to which we wish to connect. In the case of a component, we must always specify the port (which is 5999 in this example, as shown in Example 9-8). The same constructor can be used to create a client connection to the Jabber server, in which case a default port of 5222 (the standard port for client connections) is assumed if none is explicitly specified. The localname argument is used to specify the component's name, which in this case is rss.qmacro.dyndns.org. In the same way that a default port of 5222 is assumed if none is specified, a default stream namespace of jabber:client is assumed if no ns argument is specified. Since we want to connect as a component using the TCP sockets connection method, we must specify the appropriate namespace: jabber:component:accept.

This constructor call results in a stream header being prepared; it looks like the one shown in Example 9-12.

The actual connection attempt, including the sending of the component's stream header, is done by calling the connect() method on the connection object in $c:

unless ($c->connect()) { die "oops: ".$c->lastError; }

This will return a true value if the connect succeeded (success is measured in whether the socket connection was established and whether the Jabber server sent a stream header in response). If it didn't succeed, we can retrieve details of what happened using the lastError() method.

We're connected. Before performing the authenticating handshake, we're going to do a bit of preparation:

$SIG{HUP} = $SIG{KILL} = $SIG{TERM} = $SIG{INT} = \&cleanup;

The idea is that the component will be run and stopped only in certain circumstances. If it is stopped, we want to clean things up before the script ends. Most importantly, we need to make sure the registration data is safe, but also we want to play nicely with the server and gracefully disconnect. This is done in the cleanup() function.

Preparing the RSS event function and element handlers

Jabber::Connection offers a simple way of having a function execute at regular intervals. It avoids the need for setting and resetting alarms. It is the register_beat() function:

debug("registering RSS beat"); $c->register_beat(1800, \&rss);

Calling the register_beat() method takes two arguments. The first argument represents the interval, in seconds. The second is a reference to the function that should be invoked at each interval. Here, we're saying we want the rss() function called every 30 minutes (1800 seconds).

Most of the traffic relating to the component will be the headline messages emanating from it. However, we are expecting incoming IQ elements, particularly for registration in the jabber:iq:register namespace. We've also already mentioned that it's customary for components to honor basic "administrative" queries such as version checks. So the list of calls to the register_handler() method here reflects what we want to offer in terms of handling these IQ elements:

debug("registering IQ handlers"); $c->register_handler('iq',\&iq_register); $c->register_handler('iq',\&iq_version); $c->register_handler('iq',\&iq_browse); $c->register_handler('iq',\&iq_notimpl);

Whereas with Net::Jabber's SetCallBacks() function and with Jabberpy's setIqHandler() method we specify a single function to act as a handler for incoming <iq/> elements, we can specify as many handlers as we want for each element type with the register_handler() method in Jabber::Connection.

The first argument refers to the element name (the name of the element's outermost tag), and the second refers to a function that will be called on receipt of an element of that name. Each of the handlers for a particular element will be called in the order they were registered. So when an <iq/> element is received over the XML stream, Jabber::Connection will dispatch it to iq_register(), then to iq_version(), then to iq_browse(), and then to iq_notimpl(). That is, unless one of those handler functions decides that the element has been handled once and for all and that the dispatch processing for that element should stop there. In this case, that handler simply returns a special value (defined in Jabber::NS), and the dispatching stops for that element. The handlers can also cooperate, in that the dispatcher will pass whatever one handler returns into the next handler in the list and so on, so that you can effectively share data across handler events for a particular element, building up a complex response as you go.

This contextual response chain model works in a similar way to how the mod_auth_* authentication modules work in the JSM. Each one that wishes to express its interest in authenticating a user adds its "stamp" to the response to an IQ-get in the jabber:iq:auth namespace, before that response is returned to the client.

{{Note|Indeed, the author of the Jabber::Connection library has taken the (heart)beat idea—the handler chain idea and even the low-level NodeFactory mechanisms—directly from the JSM and the server libraries, in homage to the Jabber server's classic design.

</code>

==== Authenticating handshake and launch of main loop ==== Once we've set up the handlers, we're ready to make the authenticating handshake. This is simply a call to the auth() method:

$c->auth('secret');

It takes either one or three arguments, depending on whether the authentication is for a client or a component. Jabber::Connection decides which authentication context is required by looking at the specified (or default) namespace in the connection constructor call. As we specified the namespace jabber:component:accept, the auth() method is expecting a single argument, the secret specified in the <secret/> tag of the component instance definition. The auth() method performs the message digest function and sends the <handshake/> element.

It's now appropriate for us to "launch" the component, with the start() method:

$c->start;

This is the equivalent of the MainLoop() method in Perl's Tk library and is a method from which there's no exit. Calling start() causes the connection object to perform an endless loop, which internally calls a process() method on a regular basis, receiving, examining, and dispatching elements received on the XML stream. It also starts and maintains the heartbeat, to which the register_beat() method is related.

{{Note|If you wish to have more granular control over your script, you can use the process() function directly, just as you would with the Net::Jabber and Jabberpy libraries. Be aware, however, that a heartbeat is maintained only in the context of the start() method.

</code>

==== Handling registration requests ==== The first of the handlers defined for <iq/> elements is the iq_register() function. We put it first in the list, as we consider receipt of <iq/> elements in the jabber:iq:register namespace to be the most common. We want this function to deal with the complete registration conversation. This means it must respond to IQ-get and IQ-set requests.

sub iq_register { my $node = shift; debug("[iq_register]");

The element to be handled is the primary piece of data that the dispatcher passes to a callback. The element is received by the $node variable, which is a Jabber::NodeFactory::Node object. Jabber::NodeFactory is the wrapper around the class that actually represents the elements (the nodes). Nodes are created using the Jabber::NodeFactory class. The first thing we should do is make sure it's appropriate to continue inside this function, which is designed to handle only jabber:iq:register-qualified queries. The namespace jabber:iq:register is represented with the constant NS_REGISTER, imported from the Jabber::NS module:

return unless my $query = $node->getTag(, NS_REGISTER); debug("--> registration request");

The getTag() method can have up to two arguments. The first can be used to specify the name of the tag you want to get, and the second argument can contain a namespace to narrow down the request. For example, there are two <x/> elements in this <message/> element:

<message to='dj@qmacro.dyndns.org' from='piers@jabber.org' id='2941'> <body>Let me know when you're ready to go</body> <x xmlns='jabber:x:event'><displayed/></x> <x xmlns='jabber:x:delay' from='dj@qmacro.dyndns.org' stamp='20010831T08:58:30'>Offline Storage</x> </message> We could distinguish one <x/> element from the other by specifying either the jabber:x:event or the jabber:x:delay namespace in the second argument to the getTag() function.

Although the query tag is normally placed within an <iq/> element that has the name "query," we see from Section 5.4.3.2 that it could be anything.

$node->getTag(, NS_REGISTER) This statement says "get a single child tag of the <iq/> node—regardless of its name—and qualify it with the jabber:iq:register namespace."

If we call the getTag() function in scalar context, and there is more than one tag that matches, only the first one found will be returned. If we call it in list context, all the matching tags are returned. Assuming the call is successful, the variable $query then contains the <query/> tag and all its subtags. So if we received this in $node:

RECV: <iq type='set' id='JCOM_5' to='rss.qmacro.dyndns.org' from='dj@qmacro.dyndns.org/basement'> <query xmlns='jabber:iq:register'> <text>Slashdot</text> </query> </iq>

then $query would contain a Jabber::NodeFactory::Node object that represented this bit:

<query xmlns='jabber:iq:register'> <text>Slashdot</text> </query> If the jabber:iq:register namespace doesn't qualify any of the child tags, iq_register() returns and the dispatcher calls the next handler in line (iq_version()). However, let's assume that we do have a registration IQ on our hands. The function must handle both IQ-get and IQ-set. We first deal with a potential IQ-get:

# Reg query if ($node->attr('type') eq IQ_GET) { $node = toFrom($node); $node->attr('type', IQ_RESULT); my $instructions = "Choose an RSS source from: ".join(", ", keys %sources); $query->insertTag('instructions')->data($instructions); $query->insertTag('text'); $c->send($node);

 }

The attr() method called on a node will return the value of the node's attribute of the name specified as the first argument. We test to see if IQ's type attribute is get (IQ_GET). If it is, we need to return an IQ-result as shown in Example 9-14.

Rather than create a new element from scratch, to return in response, we simply "convert" the incoming element by making necessary changes to it, turn it around and sent it back out as the response. The first thing we do is swap the values for the from and to attributes in the <iq/> tag (in $node) by calling the toFrom() function (see Section 9.3.4.11) and setting the value for the type attribute to result by calling a two-argument version of the attr() function, which turns this:

<iq type='set' id='JCOM_5' to='rss.qmacro.dyndns.org' from='dj@qmacro.dyndns.org/basement'>

into this:

<iq type='result' id='JCOM_5' from='rss.qmacro.dyndns.org' to='dj@qmacro.dyndns.org/basement'>

Notice that the from attribute is retained. This is required as the script is a component, and the response won't get stamped with one.

The instructions and an empty <text/> tag must be passed back in the response. The names of the sources are combined into a list, and an <instructions/> tag is inserted into the query node (in $query) containing the text. This is done with two method calls: the first to insertTag(), which returns a Jabber::NodeFactory::Node object that represents the newly inserted tag, and the second to data(), which inserts (or retrieves) data into (or out of) a node. The line:

$query->insertTag('instructions')->data($instructions);

could have been written as:

my $instructions = $query->insertTag('instructions'); $instructions->data($instructions);

Once constructed, the response now looks like this:

<iq type='result' id='JCOM_5' from='rss.qmacro.dyndns.org' to='dj@qmacro.dyndns.org/basement'> <query xmlns='jabber:iq:register'> <instructions> Choose an RSS source from: jonudell, slashdot [...] </instructions> <text/> </query> </iq>

This is sent using the send() method of the connection object.

If the query wasn't an IQ-get, then it might be an IQ-set:

# Reg request if ($node->attr('type') eq IQ_SET) {

   # Strip JID to user@host
   my $jid = stripJID($node->attr('from'));

   $node = toFrom($node); my $source;

In this case, the user is requesting new items for an RSS source she's specified in the <text/> field carried in the query part of the IQ-set. The user's JID can be found in the from attribute of the element, which is extracted with the attr() method. There's one thing we should do before using that JID as a key in storing that user's RSS source preferences. Look at what the JID was in earlier examples:

dj@qmacro.dyndns.org/basement As you can see, this JID has a resource attached to it. That's fine for returning a response to an IQ request, but we need something less specific. The resource part of the JID reflects the client connection of the user at the time the registration request was made. In the future, when we have an RSS item to send to her, she might be connected with a different resource. We want the RSS item to go to the right place, so we use only the user and hostname JIDs to store preferences and subsequently address the headline messages. The more generic form of the JID can be obtained by calling the stripJID() function, described later.

After swapping the from and to values, we then deal with the two different types of IQ-set requests: a request to receive a specific source or a request to cancel the registration:

# Could be an unregister request if ($query->getTag('remove')) { delete $reg{$jid}; $node->attr('type', IQ_RESULT);

   }
   # Otherwise it's a registration request for a source
   elsif ($source = $query->getTag('text')->data and
   exists($sources{$source})) { my $element = $reg{$jid};
   $element->{$source} = 1; $reg{$jid} = $element;
   $node->attr('type', IQ_RESULT);
   }

Sending a <remove/> tag in an IQ-set registration context represents a request to unregister. So we honor that by removing any trace of the user's JID from the registration hash and simply changing the type of the <iq/> element to result. Otherwise, we interpret the IQ-set as a request to subscribe to the RSS source that she's specified in the <text/> tag. We extract that source's name into $source, check that it's valid, and add a reference to the user's list of sources in the registration hash %reg. Example 9-18 shows what the registration hash looks like.

Typical contents of the registration hash

( 'dj@qmacro.dyndns.org' => { 'slashdot' => 1

                           }
 'piers@jabber.org'     => { 'jonudell' => 1 'slashdot' => 1
                           }
 ... )

In case you're wondering what's going on with the $element variable, it's because of a current restriction with MLDBM. Although it allows us to store complex structures via DB_File, those structures can't be directly manipulated, so we have to do it via a proxy variable, using $element. The value in the registration hash %reg pointed to by the $jid key is a structure. To manipulate that structure, we need to take a reference to it, in $element. The structure can be manipulated through $element and then placed back into the registration hash by assigning $element as the value corresponding to the $jid key.

Once this is done, we also mark the fact that the request was completed by setting the IQ element's type attribute to result.

Finally, it's worth telling the requester that anything else sent isn't allowed:

else { $node->attr('type', IQ_ERROR); my $error = $node->insertTag('error'); $error->attr('code', '405'); $error->data('Not Allowed');

   }

That is, if we haven't understood what the IQ-set was—it wasn't a <remove/> request nor was it a subscription to a source we recognize—we simply return it with an <error/> tag like this:

RECV: <iq type="set" id="jimAgentID657" to="rss.qmacro.dyndns.org" from="dj@qmacro.dyndns.org/basement"> <query xmlns="jabber:iq:register"> <text>banana</text> </query> </iq>

SEND: <iq id='jimAgentID657' type='error' from='rss.qmacro.dyndns.org' to='dj@qmacro.dyndns.org/basement'> <query xmlns='jabber:iq:register'> <text>banana</text> </query> <error code='405'>Not Allowed</error> </iq>

The <iq/> type is set to error to draw the client's attention to the <error/> tag. Sending an element back in error is a great example of when reusing an incoming element to build the outgoing response works very well; we don't have to reproduce what the error was, as it's already contained in what we're returning to the user.

In all of the IQ-set cases, we want to send something back:

$c->send($node);

 }
 return r_HANDLED;

Note that we also return a special value, r_HANDLED. The fact that we've got this far means that an IQ element was received and that it was a registration-related element. It's been handled, so there's no point in the other callbacks registered to handle IQ elements to get a look in. So we tell the dispatcher to stop the invocation chain for the element just processed.

==== Handling version requests ==== Now that we've seen the iq_register() function, the function to handle jabber:iq:version queries looks pretty straightforward:

sub iq_version {

 my $node = shift; debug("[iq_version]");

 return unless my $query = $node->getTag(, NS_VERSION) and
 $node->attr('type', IQ_GET);

 debug("--> version request");

 $node = toFrom($node); $node->attr('code', IQ_RESULT);
 $query->insertTag('name')->data($NAME);
 $query->insertTag('version')->data($VERSION);
 $query->insertTag('os')->data(`uname -sr`); $c->send($node);

 return r_HANDLED;

Just as we check for whether the element is appropriate to handle in iq_register(), we do here, too, this time looking for an IQ-get with a query child tag qualified by the NS_VERSION (jabber:iq:version) namespace, which we grab in $query.

Setting the <iq/>'s type to result and flipping the addresses, we then just have to add <name/>, <version/>, and <os/> tags to the query child with appropriate values, to end up with a response like the one shown in Example 9-16.

If we do this, we deem the IQ to have been handled and return the special r_HANDLED value, as before, to stop the dispatching going any further for this element.

==== Handling browse requests ==== Next in line to handle the incoming <iq/> element is the iq_browse() function. Of course, if we've already handled the element, iq_browse() won't even get a shot at responding. But if it did, it would proceed along similar lines to the iq_version() function:

sub iq_browse {

 my $node = shift; debug("[iq_browse]");

 return unless my $query = $node->getTag(, NS_BROWSE) and
 $node->attr('type', IQ_GET);

 debug("--> browse request");

 $node = toFrom($node); $node->attr('type', IQ_RESULT); my $rss =
 $query->insertTag('service'); $rss->attr('type', 'rss');
 $rss->attr('jid', $ID); $rss->attr('name', $NAME);
 $rss->insertTag('ns')->data(NS_REGISTER); $c->send($node);

 return r_HANDLED;

The only real difference is that we want this function to handle IQ-gets in the jabber:iq:browse namespace and return a browse result. Browsing will be discussed in greater detail in Section 10.3 in Chapter 10. For now, though, let's focus on returning a top-level browse result that reflects what might be returned if a similar browse request were made of the JSM, as described in Section 6.2.5 in Chapter 6. Example 9-19 shows what iq_browse() will return.

The RSS news agent responds to jabber:iq:browse requests via iq_browse()

RECV: <iq type="get" id="browser_JCOM_2" to="rss.qmacro.dyndns.org"> <query xmlns="jabber:iq:browse"/> </iq>

SEND: <iq id='browser_JCOM_2' type='result' to='dj@qmacro.dyndns.org/winjab' from='rss.qmacro.dyndns.org'> <query xmlns='jabber:iq:browse'> <service jid='rss.qmacro.dyndns.org' type='rss' name='RSS News Agent'> <ns>jabber:iq:register</ns> </service> </query> </iq>

==== Dealing with other requests ==== There are untold IQ elements that could be sent to the component. While it would be possible just to ignore them, we ought to do something and at least respond with something like "not supported." For that, we have iq_notimpl() as a catch-all. If the dispatcher manages to make its way to here, we know that the <iq/> element is not anything we recognize as wanting to respond to.

The following can be used to tell the requester that what they're asking for hasn't been implemented:

sub iq_notimpl { my $node = shift; $node = toFrom($node); $node->attr('type', IQ_ERROR); my $error = $node->insertTag('error'); $error->attr('code', '501'); $error->data('Not Implemented'); $c->send($node);

 return r_HANDLED;

As you can see, all this does is set the <iq/> type to error, switches the from and to, and adds an <error/> tag:

<error code='501'>Not Implemented</error> The modified element is sent back to the requester. This lets them know that they've requested something that just isn't there or hasn't been implemented yet.

==== The RSS mechanism ==== Now that we've set up the functions to handle incoming queries, all that's left is for us to define what happens every time the heartbeat in the Jabber::Connection loop ticks past the 30-minute mark. We registered this rss() function with the register_beat() method earlier in the script:

sub rss {

 debug("[rss]");

 # Create NodeFactory
 my $nf = new Jabber::NodeFactory;

In the IQ handlers iq-register(), iq-version(), iq-browse(), and iq-notimpl(), we avoided the need to build elements from scratch, by simply turning around the incoming request elements and making them into responses, before sending them back. Here, in the rss() function, we'll actually be building elements from scratch—headline type <message> elements to be precise. This is the reason we need an instance of the Jabber::NodeFactory.

# Go through each of the RSS sources foreach my $source (keys %sources) {

   # Retrieve attempt
   my $data = get($sources{$source});

   # Didn't get it? Next one
   unless (defined($data)) { debug("cannot retrieve $source"); next;
   }

   # Parse the RSS
   my $rss = XML::RSS->new(); eval { $rss->parse($data) };

   if ($@) { debug("Problems parsing $source"); next;
   }

The procedure in this function reflects what we described earlier in Section 9.3.1.3. Each time rss() is called, it goes through each of the sources defined in the list (%sources) and tries to retrieve it with get(), a function from the LWP::Simple library, and parse it with an instance of XML::RSS. Because XML::RSS uses XML::Parser, which dies if it encounters invalid XML, we wrap the call to the parse() method in eval.

{{Note|Ideally we'd use just one instance of XML::RSS for the whole rss() function, but XML::RSS requires us to create a new instance for every source we wish to work with.

</code>

my @items = @{$rss->{items;

   # Check new items
   debug("$source: looking for new items"); foreach my $item (@items) {

     # Stop checking if we get to items already seen
     last if exists $cache{$source} and $cache{$source} eq
     $item->{link};

     debug("$source: new item $item->{title}");

</code> Pulling the items from the RSS source into @items, we look through them, but stop looking if we come across one that we've seen previously (and stored in the %cache).

If we do have a new item to send out, we create a headline message containing the item's details:

# Create a headline message my $msg = $nf->newNode('message');

     $msg->attr('type', 'headline'); $msg->attr('from', join('@',
     $source, $ID));
     $msg->insertTag('subject')->data($item->{title});
     $msg->insertTag('body')->data($item->{description});

     my $xoob = $msg->insertTag('x', NS_XOOB);
     $xoob->insertTag('url')->data($item->{link});
     $xoob->insertTag('desc')->data($item->{description});

We use the node factory in $nf to create a new empty <message/> element, with the newNode() method. This element is built into a full-blown headline message with a jabber:x:oob-qualified <x/> extension containing the RSS item information. We can see that the call insertTag() used here has two arguments. The second is used to specify an optional namespace with which the new node (or tag) will be qualified. This call creates a Jabber::NodeFactory::Node object in $xoob that looks like this:

<x xmlns='jabber:x:oob'/> This is then embellished with the usual <url/> and <desc/> tags. What's still missing is the address information. We've specified the from; indeed taking a departure from the values we've specified for the from in the IQ responses, here we specify something slightly different with join('@', $source, $ID), which is a [user]@[hostname]-style address. For example, the source for Slashdot would be:

slashdot@rss.qmacro.dyndns.org This is mostly because it conveys more information than just the component name rss.qmacro.dyndns.org would. While the component's address would not normally be seen by a client user in the context of the IQ responses, many Jabber clients that support the headline message type will show the message sender in the headline list display. You can see this in Figure 9-8, where the From column in the headline list clearly shows the RSS source from which the item originated. This has a little bit of future for the component built in, too. If we wanted to extend the component for more interaction with the clients, we could have the client send a message to the [RSS source]@[componentname] JID and, on receipt, the component would immediately have context information on which source the message was about, without the client user having to do anything other than specify a JID.

{{Figure|title=Jarl's headline display window|image=0596002025-jab_0908.png</code> Now that we've built the headline message, which looks like the one in Example 9-9, we can fire it off to each of the users who have registered for that RSS source:

# Deliver to all that want RSS headlines foreach my $jid (keys %reg) {

       my $registration = $reg{$jid};

       if (exists($registration->{$source})) { $msg->attr('to',
       $jid); debug("delivering to $jid"); $c->send($msg);
       }

     }

The first time we encounter an RSS source, we won't have any record of a "last seen" item in the %cache. So we avoid flooding people with all the items of a new RSS source by jumping out of the item loop if there's no cache info:

# Prevent all items counted as new the first time around last unless exists($cache{$source});

   }

Finally, we make a mark in the cache for the latest item we've encountered, to prepare the script for the next feed:

# Remember the latest new item $cache{$source} = $items[0]->{link};

 }

==== The cleanup() function ==== The cleanup() function is called if an attempt is made to shut the script down. It unties the registration hash, ensuring no data is lost, and disconnects from the Jabber server:

sub cleanup {

 debug("cleaning up"); untie %reg; $c->disconnect; exit;

==== Helper functions ==== Any script over a certain size is likely to have some helper functions, and our RSS news agent is no exception. Here we have the function to switch the from and to attribute values of a node (toFrom()), the function to remove the resource from a JID (stripJID()), and something not much better than a debugging-style print statement:

sub toFrom { my $node = shift; my $to = $node->attr('to'); $node->attr('to', $node->attr('from')); $node->attr('from', $to); return $node; }


sub stripJID { my $JID = shift; $JID =~ s|/.*$||; return $JID; }

sub debug { print STDERR "debug: ", @_, "\n"; }

=== Further Ideas === There's only so much that can be included in a demonstration script. There's plenty of room for improvement, even if you don't count rewriting it all from scratch. For example, you could store registrations in a SQL database or, alternatively, use the Jabber server's own xdb component. More importantly, a static list of RSS sources is rather restrictive. How about allowing the user to register their own URLs? Or building an administrative mode that accepts a special IQ from certain JIDs, with which the RSS source list can be maintained?

The browsing response function would be an ideal candidate to be expanded upon. Another level of browsing could be added that would return browse items that reflect the specific user's RSS source registrations. It's probably worth exploring the power of addressing the component to include the RSS source, to extend the interactive facilities. For example, you might want to have the script accept and act upon messages sent to the component's JID, which includes a username portion representing the RSS source.

== A Simple Headline Viewer == Let's close this chapter on a lighter note by building a complementary script that we can use to keep an eye on the headlines coming in from the RSS news agent. We'll call this program hlv, for "HeadLine Viewer." Basically, we want to build something that allows the user to watch RSS headlines as they scroll by in a window that doesn't take up a lot of screen real estate.

In the recipes so far, we've had great success building solutions that make use of off-the-shelf clients. But it's time to buck the trend, indeed to make another point. While you can arrange the features of your Jabber-based applications in the direction of standard clients, to take advantage of the installed base of Jabber clients, if you do want to create a client that works differently, a client that fits your needs exactly, then go ahead—it will be surprisingly straightforward. The mantra of "server side complex, client side simple" (with apologies to George Orwell) is there to help us. What's more, we can put into action an idea expressed earlier in the book (in Section 2.7):

A Jabber client is a piece of software that implements as much of the Jabber protocol as required to get the job done.

that the information is going to be delivered to us in headline-type <message/> elements, why have the viewer client understand or deal with anything else? To implement a Jabber solution, we pick and choose the parts of the protocol that make sense in the context of that solution. If you want to transport RPC calls in data payloads between two endpoints, why bother with roster management or rich-text chat facilities? If you just want to join a conference room to talk with the other room occupants, why bother with threaded one-on-one conversations? If you need routing and presence capabilities to have your oven know when you've arrived home, why engineer anything else?

What we're going to write here is a simple headline viewer. Nothing more, nothing less. It will know the tiniest thing about presence—as the headlines come in as <message/> elements, it will need to announce its availability to the JSM. We're going to build a Jabber client that will have a session context with the JSM, so we need to tell the JSM that we're available when the client starts. Otherwise, the headlines will be queued by the store-and-forward mechanism for delivery the next time we're available.

We'll leave registration to the RSS news agent to another client that knows about agents (services) and can interact in a jabber:iq:register kind of way. Again, the "one size fits all, one client for everything" philosophy doesn't always have to apply; different features of different programs can get the job done. So while the headline viewer will receive, understand, and display the message headlines, we'll use WinJab or even JIM to manage the RSS source subscriptions. Figure 9-9 illustrates the process of registration with the RSS news agent component, using JIM.

{{Figure|title=Registering with the RSS News Agent, using JIM|image=0596002025-jab_0909.png</code> The suggestion of JIM as a client to complement or make up for the lack of features in the headline viewer is deliberately ironic. The role JIM plays is to provide support for core Jabber client features, of which headline messages are not considered to be a part. So while JIM can interact with services and register (and unregister—which will send the <remove/> tag in the query, as described in Section 9.3.4.5), it doesn't handle headline-type messages, which is perfectly fine. Our headline viewer won't handle chat or normal messages—it's not supposed to.

It's worth pointing out that another reason the headline viewer client can remain simple is because the RSS news agent will be doing all the hard work. Unlike other (non-Jabber) headline viewer programs, hlv depends upon the RSS news agent. It's that component that will maintain the list of RSS sources. It's also that component that will retrieve those sources at regular intervals and check for new items. All we have to do is sit back and have those new items pushed to us, at which point the client has to make a slight effort to insert the details of those new items into the viewer display.

=== What the Headline Viewer Is Going to Do === The headline viewer, shown in Figure 9-10, has a scrollable area where headlines are displayed. We can clear that area, or select a headline and call up a web browser to fetch the story by passing the URL to it.

{{Figure|title=The Headline Viewer client|image=0596002025-jab_0910.png</code> It's also nice and small, visually and in the amount of code we're going to have to write. We connect to the Jabber server, set up a handler for the incoming headline messages, build the display, send our availability, and sit back and watch the news roll in.

Actually, we need to say a few things about the "sitting back" bit. We know that Jabber programming implies an event model. For this example, we're going to use Tk, a widget library for building GUI applications, with bindings for many languages. Tk itself has an event model, which in many ways reflects Jabber's. Table 9-3 shows how Jabber and Tk relate to each other in this programming model.

Having one program governed by two independent event loops is not what we want to try to achieve. We want Jabber's and Tk's event models to cooperate. This is achievable by using a master/slave relationship between the two models. Tk's repeat() method can be used to invoke a function that calls the Jabber library's process() method. We can hand over control to Tk with MainLoop() and know that the Jabber event model will get a look in because of the Tk event callback we've defined with repeat().

=== The hlv Script === The hlv script, shown in Example 9-20, uses Perl with the Net::Jabber library.

The hlv script, written in Perl

use Tk; use Net::Jabber qw(Client); use strict;

use constant SERVER => 'gnu.pipetree.com'; use constant PORT => 5222; use constant USER => 'dj'; use constant PASSWORD => 'secret'; use constant RESOURCE => 'hlv';

use constant BROWSER => '/usr/bin/konqueror';

my @headlines; my @list;

my $connection = Net::Jabber::Client->new();

$connection->Connect( hostname => SERVER, port => PORT, ) or die "Cannot connect ($!)\n";

my @result = $connection->AuthSend( username => USER, password => PASSWORD, resource => RESOURCE, );

if ($result[0] ne "ok") { die "Ident/Auth with server failed: $result[0] - $result[1]\n"; }

$connection->SetCallBacks( message => \&handle_message );

my $main = MainWindow->new( -title => "Headline Viewer" ); $main->geometry('50x5+10+10'); $main->repeat(5000, \&check_headlines);

             },
           );

             },
          );

 }

 my ($oob) = $msg->GetX('jabber:x:oob'); push @headlines, { link
 => $oob->GetURL(), title => $msg->GetSubject(),
                  };

=== Reviewing the hlv Script Step by Step === The script starts with the declarations of the libraries we're going to use, along with some constants:

use Tk; use Net::Jabber qw(Client); use strict;

use constant SERVER => 'gnu.pipetree.com'; use constant PORT => 5222; use constant USER => 'dj'; use constant PASSWORD => 'secret'; use constant RESOURCE => 'hlv';

use constant BROWSER => '/usr/bin/konqueror';

The hlv script will connect to Jabber as a client, so we need to specify that in the use statement to have the appropriate Net::Jabber modules loaded. We're going to be connecting to the Jabber server at gnu.pipetree.com, although, as we said, the RSS news agent might live somewhere else. It just so happens that in this scenario, there's a reference to the component in gnu.pipetree.com's JSM <browse/> section, so we can carry out registration conversations with it (using JIM, for example).

If the Fetch button is clicked when an item in the list is selected (see Figure 9-10), we want to jump to the story by launching a web browser. The constant BROWSER used here refers to the browser—in this case, Konqueror, the browser of choice for KDE users—on the local machine.

my @headlines; my @list;

We declare two arrays: @headlines, which we'll use to hold the items as they arrive contained in the headline <message/> elements on the XML stream, and @list, to hold the URLs that relate to those items in @headlines.

After connecting to and authenticating with the Jabber server, we set up the callback to take care of incoming <message/> elements (this is very similar to the way the coffee monitor script connects and authenticates in Section 9.2.3.5):

my $connection = Net::Jabber::Client->new();

$connection->Connect( hostname => SERVER, port => PORT, ) or die "Cannot connect ($!)\n";

my @result = $connection->AuthSend( username => USER, password => PASSWORD, resource => RESOURCE, );

if ($result[0] ne "ok") { die "Ident/Auth with server failed: $result[0] - $result[1]\n"; }

This is the handle_message() function:

$connection->SetCallBacks( message => \&handle_message );

Now it's time to build the GUI. We start by creating a main window, giving it a title and geometry, and establishing the cooperation between the two event models with the repeat() method:

my $main = MainWindow->new( -title => "Headline Viewer" ); $main->geometry('50x5+10+10'); $main->repeat(5000, \&check_headlines);

The repeat() function will arrange Tk's main event loop to hiccup every five seconds (the first argument is measured in milliseconds) and call the check_headlines() function.

Next, we build a frame to hold three buttons—Clear, Fetch, and Exit—and a scrollable list to contain the titles as they're received:

# Button frame my $buttons = $main->Frame(); $buttons->pack(qw/-side bottom -fill x/);

Defining the buttons, one at a time, brings our attention to the Tk UI event model, in that we define the handlers using the -command argument of the Button() method. The handlers' jobs are quite small, so we can get away with writing them inline:

# Clear button my $button_clear = $buttons->Button( -text => 'Clear', -underline => '0', -command => sub { @list = (); $list->delete(0, 'end')

             },
           );

If called, the Clear button clears the scrollable display by calling the delete() method on the $list object and emptying the corresponding array of URLs.

The Fetch button extracts the URL from the highlighted item in the scrollable list using the curselection() method to retrieve the index value. The @list array is then used to look up the URL by launching the browser (in this case, Konqueror) in the background. Many browsers accept a URL as the first argument; if your choice of browser doesn't, you'll need to modify this call slightly.

# Fetch Button my $button_fetch = $buttons->Button( -text => 'Fetch', -underline => '0', -command => sub { system( join(" ", (BROWSER, $list[$list->curselection], "&")) )

             },
          );

If clicked, the Exit button uses a destroy() function to, well, destroy (or close) the main window. This causes Tk's main event loop to come to an end, passing control back to the statement in the script following the one with which that main event loop was launched (with MainLoop()):

# Exit button my $button_exit = $buttons->Button( -text => 'Exit', -underline => '0', -command => [$main => 'destroy'], );

Having created all the buttons and packed everything into the window with the pack() method:

$button_clear->pack(qw/-side left -expand 1/); $button_fetch->pack(qw/-side left -expand 1/); $button_exit->pack(qw/-side left -expand 1/);

$list->pack(qw/-side left -expand 1 -fill both/);

we announce to the JSM that we're available:

$connection->PresenceSend();

All that remains for us to do is start Tk's main event loop. We include a call to the Net::JabberDisconnect() method for when the Exit button is clicked and control returns to the script, so we can gracefully end the Jabber connection:

MainLoop();

$connection->Disconnect; exit(0);

We defined the check_headlines() function as the function to invoke every five seconds:

sub check_headlines { $connection->Process(1); while (@headlines) { my $headline = pop @headlines; $list->insert(0, $headline->{title}); unshift @list, $headline->{link};

 }

To check for any messages that have arrived on the XML stream, we can call the Process() method on the connection object. If there are any messages waiting, the handle_message() function is called to handle them:

sub handle_message { my $msg = new Net::Jabber::Message($_[1]); return unless $msg->GetType eq 'headline';

 my ($oob) = $msg->GetX('jabber:x:oob'); push @headlines, { link
 => $oob->GetURL(), title => $msg->GetSubject(),
                  };

We can see fairly easily what the GetX() method does, if we remember that a headline message, complete with an <x/> extension qualified by the jabber:x:oob namespace, looks like this:

<message type='headline' to='dj@qmacro.dyndns.org'> <subject>JabberCon Update 11:45am - Aug 20</subject> <body>JabberCon Update - Monday Morning</body> <x xmlns='jabber:x:oob'> <url>http://www.jabbercentral.com/news/view.php?news_id=998329970& lt;/url> <desc>JabberCon Update - Monday Morning</desc> </x> </message>

It returns, in list context, all the <x/> elements contained in the element represented by $msg that are qualified by the jabber:x:oob namespace. We're expecting there to be only one, which is why we plan to throw all but the first array item away with the ($oob) construction. After the call to GetX(), the object in $oob represents this part of the message:

<x xmlns='jabber:x:oob'> <url>http://www.jabbercentral.com/news/view.php?news_id=998329970& lt;/url> <desc>JabberCon Update - Monday Morning</desc> </x>

The item's details, including the URL and title, are pushed onto the @headlines list, and the headline-type message-handling function has done its job. Control passes back to the check_headlines() script to immediately after the call to the Process() method.

The handle_message() function may have been called multiple times, depending on how many elements had arrived, so the @headlines array might contain more than one item. We run through the array, sending off each headline in turn, inserting the title into the scrollable list object and the URL into the corresponding position in the @list array:

$list->insert(0, $headline->{title}); unshift @list, $headline->{link};

That's really all there is to the code. The features are minimal, but the hlv script gets the job done. There's just enough programming to be able to receive and display RSS items.

The "just enough" philosophy demonstrated in this recipe can be appropriate in non-GUI applications of Jabber as well. Taking away the visual element of hlv to leave the mechanism for receiving and understanding jabber:x:oob information, we're left with a Jabber-aware "stub" that can be put to many uses; collection and aggregation of news stories comes immediately to mind. Along the same lines, we can see that the way we gave presence to the coffee pot in Section 9.2 was just a melding of a stub, which understood Jabber presence, with a loop mechanism that connected to the MINDSTORMS device.

It's possible to build extremely powerful solutions using Jabber's protocol and the building blocks that the protocol represents (described in Chapter 6), without having any relation to a user interface. We'll see such a solution in Section 10.2.

</content></chapter>