[Twisted-web] Copying Nevow features into Twisted Web (was Re: Status of divmod-dev list?)

Yaroslav Fedevych jaroslaw.fedewicz at gmail.com
Tue Dec 21 07:46:41 EST 2010


On Mon, Dec 20, 2010 at 5:43 PM, Werner Thie <werner at thieprojects.ch> wrote:
> On 12/20/10 4:19 PM, Colin Alston wrote:
>> Primarily there is a great deal of bootstrapping. Just create a simple
>> web-app (without something like Methanal or Mantissa, which I can't even
>> understand) that does Athena and see how much code you're rewriting
>> every time, it's quite insane.
> My boilerplate (me working with guys mostly doing PHP and MySQL) let's me
> start up a new project in less than an hour. Doing this for a living I
> see the problem not really in starting up projects fast, but in
> finishing them fast, because only finished projects are paying...
>

The problem is that your boilerplate wasn't a part you wrote in an
hour, and it's not something everyone can pick up and use.

>> Also if you're not using Twisted plugins (which there are many reasons
>> not to want to, like using shared state reactors with multiple TCP
>> services, or again Freeze) the pain goes up double.
> I use only a single plugin in nevow letting me bootstrap easily and
> providing
> a clean mechanism to switch automagically between source/compressed
> deployment version. All the deployment is handled via SVN.
>

Again, I use plugins quite a lot (after all, they exist for a reason),
and I don't use SVN. The point is that I write a bunch of servers and
the web-exposed part constitutes 10%, if not less, in a single setup.
Hence, all the divmod'esque stuff with Mantissa and friends is an
overkill, but trying to do everything on top of plain t.web leaves a
bad feeling of reimplementing Nevow.

>> Essentially boiling Athena down to a page mixin class with standard js
>> bootstraps, and/or separate fragment mixin and making all pages derive
>> the necessary Athena object triggers or even putting that into flatten.
> Might be that there's need for a 'kiddy' version, letting the average
> interested user who does not have the time to look into a framework and
> its concepts and wants to whip up a page in less than 30 seconds.
>

The truth (maybe sad one) is that there are still areas where a simple
generated HTML with AJAX interactions sprinkled here and there is the
optimal thing. And sometimes all you want is throw away your old PHP
frontend to some pythonic monster and bring it all together to a
single code base, without much rethinking of the interaction model.

-- 
Engineer : How do I do it?
Economist : How much will it cost?
Twisted Developer: But does it block?



More information about the Twisted-web mailing list