[Twisted-web] twisted.web2+Cheetah+FormKit

[glyph at divmod.com]

On Thu, 19 Jan 2006 13:15:47 -0800, Ed Suominen <general at eepatents.com> wrote:
>glyph at divmod.com wrote:

>Unfortunately,  even after all this time has passed, there still doesn't
>seem to be any solid footing for maintaining a Twisted-based web server.

This is definitely a big problem.  The fact that web2 hasn't stabilized yet is actually an indication that things are going to get better.  We are trying to minimize the number of half-supported revisions out there, so that when web2 is released as "stable", there is a clear, defined migration path.

I understand your frustration, and to some extent, I even share it - believe me, Divmod's life would be a lot easier if web2 were totally stable and productionized and we could just move Nevow over to it.

>there's not exactly a shortage of Python frameworks.

That's very true.  Unless you have a need for Twisted in your web application, there's no particular reason you need to use it.  There's a certain irrational exuberance that follows Twisted and related projects around, which sadly encourages people to develop apps on top of it that have no compelling reason to be on top of it, rather than first contributing the development and maintenance effort required to productionize whatever subproject is being used.

It's important to get some perspective, though.  Crummy as the current web situation is in Twisted, there is an INSANE amount of stuff in the TwistedVerse right now even if you don't look beyond the small sphere of the Twisted project and the Divmod repository.  There is a group of maybe two dozen people (and that's being really optimistic) maintaining a list of project that competes, at least partially, with each of the following projects that corresponds to a category of software:

apache (web/web2)
turbogears (nevow/mantissa)
asterisk (sine)
sqlobject (axiom)
skype (shtoom)
jabberd (words)
gaim (words.im, MOST, etc)
bind (names)
wordpress (quotient, eventually hyperbola)
exim (xquotient)
nnptd (twisted.news)
cyrus (twisted.protocols.imap4, quotient)
ACE (twisted.internet)
openssh (twisted.conch)
ncurses (twisted.conch.insults)
docbook (twisted.lore)
ORBit (twisted.pb)
inetd (twisted.runner)
jUnit (trial)

Then you have projects like Vertex and Radical, for which there _is_ no meaningful competitive space yet, because they're so unique.

Granted, in each of these categories, most of the alternatives to Twisted are more full-featured than the Twisted project at this point.  However, this is an _absolutely insane_ amount of code to have been written by the tiny number of people who work on Twisted on a regular basis.  There is no other project which even comes remotely close to having this much functionality available for integration into a single paradigm (and not only that, a single process, if you want).  On each of those competitive projects, there is a comparable number of distinct developers to the entire Twisted+Divmod team.  It's overwhelming, and maintenance slips sometimes.

Overall, we need HELP.  If you're just a user, every project needs some help.  Find some bugs in the Twisted or Divmod bugtracker; there are plenty related to web2 or Nevow.

>That said, I continue to be fascinated and impressed by Twisted itself
>and have high hopes for it.

Thanks.  It's always nice to hear this.  I guess I just probably explained why that is true :).