// JavaScript
var faye = require('./faye');
var comet = new faye.NodeAdapter({mount: '/comet'});

# Ruby
require 'faye'
comet = Faye::RackAdapter.new(:mount => '/comet')

<script type="text/javascript" src="/comet.js"></script>
<script type="text/javascript">
  CometClient = new Faye.Client('/comet');
  CometClient.connect();
</script>

// I can publish a message ...
CometClient.publish('/my/channel', {msg: 'Hello world!'});

// .. and you can pick it up
CometClient.subscribe('/my/channel', function(message) {
  alert(message.msg);
});

16 thoughts on “Faye: a Comet client and server for Node.js and Rack”

1. That is exactly the piece of code I was looking for! Thank you!

2. Ajaxian » Faye: Bayeaux protocol Comet server for Node.js

3. Hi, can you use Faye to connect to itself from within Node.js and subscribe some channels? I’d like to monitor some channels to keep a backlog of messages in memory, and serve the backlog to clients through another URL.

4. Kenneth, I’m working on a set of server-side clients at the moment. For Node, I’m taking the browser client and adding new transports for Node’s HTTP API and for in-process client-server communication. The Bayeux protocol allows in-process clients special access to server state so it’s a little work to implement this. Once I’m happy with it I need to backport the client to Ruby before getting a release out.

   You can track progress on this branch: http://github.com/jcoglan/faye/tree/clients

5. Hello James, have looked through commit history and it seems to me that you did not yet implemented a way to invoke publish() on server side yet, am I right?

6. I have implemented server-side clients that can publish and subscribe, they’ve just not been released yet. Hopefully I’ll get them released this week, in the meantime there’s information in the README on GitHub, for example under Node you’d do:

   // Remote client
   client = new Faye.Client('http://example.com/faye');
   
   // Local client
   server = new Faye.NodeAdapter(options);
   client = server.getClient();

   And use the client just like you would in a browser.

7. Works flawlessly so far, great job, thank you!

8. Nice piece of work! However it’d be nice if there were some support for error handling. ‘Looks like your XHR transport code quietly ignores a variety of error cases.

   I’m looking specifically at the case where the server gets restarted during a long-poll request. In this case, there’s no response and the Faye clients simply stop polling. At a minimum, there should be a way to configure the behavior in this case (e.g. “retry every 10 seconds”)

   Also, I’d prefer to have a way to adjust the polling frequency of the Faye.XHR readystate check. (10ms strikes me as a bit aggressive.)

   … oh, and also, you might consider doing this:

   if (self._xhr.readyState !== 4) setTimeout(handleState, 10)

   … instead of using setInterval. It’s simpler, and avoids potential issues with multiple setInterval calls getting queued up if CPU performance lags for whatever reason.

9. make that “return setTimeout(…)”

10. Robert, error handling was on my to-do list and you’ll be happy to know it just got bumped. I just made a commit that reconnects a client with the server, including maintaining its channel subscriptions, if the server goes away for whatever reason. It does this by setting a time limit within which it expects to get a reply to connect(), since the various connect transports don’t all allow for a robust means of error detection on the request.

    This is a first attempt and certainly the timeout needs to be able to be tuned, but let me know what you think.

11. Great work. Its perfect for a project that I was trasnlating from c# to java, but this is much better.

    Have a question though. In, for example, cometd, on the client (in a browser), the comet client connects and waits until the connection is confirmed before a subscription can be set. Is there a way of knowing when the conneciton is succesfully with faye?

    cheers and great work.

12. Camilo: You can, but you don’t need to worry in this case since Faye makes sure it has a valid connection before sending subscriptions. If you need to make sure a connection is present before running some code, you can use the connect() method with a callback, which is what Faye does internally:

    myFayeClient.connect(function() {
        // code goes here
    });

    connect() simply makes sure there is a connection before running the callback function.

13. Quora

14. Are there any good tutorials for building a Comet application using Ruby on Rails? - Quora

15. Comet or Long Loop message pattern – put an end to polling! | Allan's Ramblings

16. » Comet or Long Loop message pattern – put an end to polling! Allan's Ramblings