[Twisted-Python] Noob alert - Looking for a framework for creating a long polling service

ssteinerX@gmail.com ssteinerx at gmail.com
Sat Apr 3 15:10:17 EDT 2010


On Apr 3, 2010, at 2:37 PM, Alex Clemesha wrote:

> On Sat, Apr 3, 2010 at 6:03 AM, Reza Lotun <rlotun at gmail.com> wrote:
>> Hi Steve,
>> 
>>> Have you continued, or do you plan to continue development where it was left off?
>> 
>> Well, it's something I've only be toying with. I can handle all the
>> specific Twisted fixes I'd like to make:
>>  - handling lost connections gracefully
>>  - proper use of twistd and daemonization
>>  - massive code cleanups
> 
> I would very much support you continuing development on Orbited.
> In fact, I believe I can help, or already have patches for most of the
> above issues you mention.  A lot of this is result of my "Hotdot" project,
> "Create realtime webapps using Django + Orbited + Twisted":
> http://github.com/clemesha/hotdot
> 
> Specifically, see here for Orbited and twistd working together:
> http://github.com/clemesha/hotdot/blob/master/server.py
> 
> and see the below for patches that improve handling lost connections gracefully:
> http://gist.github.com/256582
> 
> Finally, it would be awesome if you put up your improvements on Github
> (just a suggestion :-),
> then I can easily fork and starting adding improvements, and you can
> pull and we can move this forward.

I had worked with your hotdot stuff a while ago, and I think I had even sent you some django csrf patches for Django 1.2 compatibility (it was a while ago...)

Back then, I had started on refactoring the one-big-javascript file in Orbited a few months ago but, at the time, was assured that 0.8 was "just around the corner", so I left it alone.

Well...now I have need and, since Athena seems to be tightly tied to Nevow, cometd is itself, websocket is websocket, so I'm looking again.   Since Orbited seems to have gone into "unmaintained mode", I agree, it may just be time for a fork.

> p.s. I haven't found many issues with the frontend JavaScript code, but if you
> have specific issues that you could point out, I'd love to try to help.

Mostly the problems I had with it were the lack of doc, and the all-in-one 2700 line .js file, and the complete lack of unit tests of any type.

>> The only problem is that I'm not that knowledgeable about the
>> Javascript browser hacks they've employed - their Orbited.js is more
>> than half "magic" to me (probably due to me not having an opportunity
>> to work on a large js project). It's something that I'd like to get
>> into, if someone is willing to take the lead on the javascript side.
>> It could potentially be as simple as slotting in js.io
>> (http://github.com/mcarter/js.io), which Orbited.js has apparently
>> become, and which appears to have been the plan for Orbited 0.8.

The js.io site documentation link says "Sorry, we don't have documentation yet", and the doc directory on github has a powerpoint file and an echo server so I'm not sure that's going to help so much.


>>> They all seem to have gone off to play with node.js instead.
>> 
>> I don't *really* blame them, considering most of the project's merit
>> lay in the Javascript realm, and node.js is basically a (less
>> featurful) version of Twisted for Javascript (of course with the added
>> bonus of using a heavily JITed and optimized Javascript VM).
>> 
>>> Very cool, I hadn't seen that before.
>> 
>> Yes, very neat. Haven't done much with it, but I think it's a real
>> winner when you're doing something simple like long-polling/
>> broadcasting from a queue.

Well, if there's interest on reviving Orbited either as itself, or as a new project with some code from Orbited as well as some of Alex's stuff, I'm in.  

S




More information about the Twisted-Python mailing list