(OFFTOPIC: It seems, I got it wrong with my previous question about APE and chat. The problem is elsewhere)
We are looking for a free Comet server which is also free for commercial use. The server must support the following:
- basic publish/subscribe functionality, no fancy features needed;
- supports publishing messages from any type of client though some simple message format (just send some data in some certain format to some certain URL, and server sends it out to all the subscribers);
- can identify connected user (subscriber) by our supplied ID, so the user always receives the data he subscribed for, even if he has opened multiple browsers or using multiple tabs;
- does not force keep-alive connections for message publishing so PHP script can just publish and die();
- it would be great if the server could itself call some URL if some certain message is received from a publisher (like running a PHP script and passing some data);
- sane subscription channel implementation to avoid JSONP or tricky DNS subdomain configurations;
- Javascript API libraries would also be great, but we could also use jQuery - if only the server supports a simple messaging protocol;
- a simple PHP API for publishing would be a great bonus;
- stable and robust;
- up-to-date documentation and support forum or active newsgroup/IRC.
We looked at a Lightstreamer but it does not support a text protocol in a free version. Liberator is not free for commercial use, free editions of StreamHub and WebSync are limited to 10 users.
APE engine has some nasty problems: http://www.nulldevice.de/2010/09/ape-ajax-push-engine/
Ngnix with NGiNX_HTTP_Push_Module seems good but it is at beta stage and requires manual compilation.
We are a young and small company, specializing mostly on web browser games (no Flash) so we cannot afford expensive solutions, and also the commercial solutions (Lightstreamer, Liberator) have many features which we will not use anyway. We have only PHP/Javascript developers for now. We could also handle some Java / C++ / C# server API but not Perl/Python. I guess 1000 simultaneous users for our planned application would be the most optimistic value we can hope for now, so scaling problems is not an issue for now.