[Twisted-web] Good ROM tool ?

[Andrea Arcangeli]

On Tue, Jan 17, 2006 at 06:02:18PM +0100, Valentino Volonghi wrote:
> Of course IMHO this is not enough of a reason to avoid nevow. First of all
> citing nevow as fragile is unfair. You said yourself that you haven't had to
> change your code since version 0.3. And now you state that it's fragile.

Formless and most of the nevow api is fragile since it has been declared
obsoleted already (web2 seems to replace a good part of it already), and
yet I already get obsolete warnings in formless code (even despite your
intention is to drop it). All this even ignoring the epsilon pointless
dependency and atop that don't matter much to me.

I've fallen in the trap once with formless, some people likely fallen in
the atop trap (thankfully I avoided it so it's only a web GUI breakage
for me), I can't take risks anymore, I've to take the obvious route
since I can't affor the cost of forking off a project, nor the risk of
keeping porting at every new breakage of the divmod APIs.

Plus nevow has lots of inefficiencies inside (I posted the profiling a
few times, most of the time is spent in the interfaces code), so going
to something lighter and faster without interfaces bloat has some
advantage too, even if it means writing more code. I'll let you know if
performance changes are visible.

Changes like from rend.notFound to Null (like the porting from nevow
to web2 of some of my code) are not a problem for me. The problem for me
is having to rewrite from scratch stuff, without any change of a
readable diff. If the diff is readable, it's quick as well to verify
there are no bugs introduced.

That's what happened with formless (and atop), so I can't let it happen
anymore, even if that means living at a lower layer and generating the
html by hand or with Cheetah (like if i was running on top of cherrpy
instead of twisted). I'll still try to use twisted async model with
Cheetah. Since I don't depend on nevow for the html generation anymore,
I can as well plug my code right on top of web2. (in the short term of
course I'll keep using nevow on twisted.web, going on top of web2 is the
mid term plan, first I'll port klive to it that is easier since there
are no forms at all and no guard too) Hopefully after web2 will start to
stabilize, its API will live long.