Re: Quixote, WSGI, next-gen hypishness, etc

On 3/9/06, Titus Brown <[email protected]> wrote:
> the last thing the Python world needs is another Web app framework ;).
> Personally, I would rather see Quixote die as a publicly traded
> framework than inflict yet another badly documented niche distribution
> on the Python community.  It's time to *kill off* Web frameworks, not
> *produce more*.

There's a difference between making a framework for framework
developers, vs making one for users.  The latter is a "new framework"
and has to be supported; the former is just a reference
implementation, which will hopefully be incorporated into Quixote or
other existing projects at a later date.  You have to have working
code if you want people to take your ideas seriously.    Or even to
know what you yourself really want.  I agree with the goal of fewer
frameworks, more interoperability, and jettisoning anything that
duplicates Paste; I've been vocal about that for a long time.  But the
last year has taught us that you have to make frameworks to kill them,
paradoxical as it sounds.  Paste had a reimplementation of Webware and
a spinoff Wareweb.  Probably nobody uses them for real sites, but I'm
sure their presence pointed out deficiences in Paste so they could be
corrected early, and the lessons from that informed the Paste-TG
integration last week.  If Quixote had a public repository, I'd just
make a branch, and nobody would say, "Oh, horrors, another
framework!", they'd say, "Oh, he's just testing some technology." and 
 "Maybe I'll use that branch if I really like it, but I wouldn't
expect the support a full-fledged project gets, and I'd have to remain
active in the Quixote community."

I'm not opposed to making a "new framework" for users, but I'm not
committed to it either.  It's way premature to decide that; it partly
depends on what direction TG and Quixote go over the next several
months, whether TG succeeds in its grand vision and settles down and
becomes more stable, whether a compatible "Quixote with batteries"
distribution is indicated, and whether I'd want to trust my own sites
to my implementation.  But I'd never make a complete old-style
framework with its own servers, session object, and other esoterica. 
At most I'd make a Paste "component" for Quixote-like applications
(which may or may not be 100% compatible with Quixote).  If that's a
framework, it's so lightweight it's not gonna make the sky fall.

I don't know what to do about the MEMS constraints on Quixote's
development.  But that also is a question that can be answered later. 
Better to get working code up and running, and then it'll be clearer
how it can best be useful.  Just like you made QWIP, knowing it would
be useful for something, but not sure exactly what.

--
Mike Orr <[email protected]>
([email protected] address is semi-reliable)