What do you want in a new web framework?

[Paul Boddie]

Cliff Wells wrote:
>
> I disagree.  Even if most of the frameworks end up being nothing more
> than research artifacts, the fact is they embody research.  Without
> research the Python web framework space will be forever relegated to
> runner-up (and probably has-been at some point).

It's perhaps surprising that even in the areas of the software industry
that are focused on selling expertise by providing services, as opposed
to selling licences for some code, that the research aspect of
developing some software, along with the knowledge gained in the
exercise, is frequently ignored in favour of some mass of code that is
gradually approaching legacy status. Familiarity with the problem
domain is extremely useful, even if it produces no long-lasting product
such as a popular Web framework, and the mere knowledge gained in
producing something that perhaps no-one else wants to use can be
important in informing future decisions about adopting (rather than
developing) suitable frameworks, if nothing else.

[...]

> Ruby on Rails doesn't come preinstalled either.  I don't think it's
> appropriate (and apparently most Linux distros currently agree) to
> install web programming frameworks by default.

I think the comment about pre-installed frameworks either confused the
issue of availability with the default package selection in some
distros, or mistook some announcement/hype about Rails on Mac OS X as
being an indication of a wider trend.

[...]

> No, the reason Rails is successful is due to being a decent, focused
> product with *great* marketing (screencasts, anyone?).

Screencasts? Perhaps, like a great showman, they draw in the punters
effectively enough, but I'd rather developers concentrate on writing
decent documentation than stuffing the pipes of the Internet up with
multi-megabyte proprietary blobs showing some individual developing
"hello world" (and still having to practise slight-of-hand in order to
make it "slick" enough).

[...]

> Also the fact that Ruby doesn't suck isn't hurting Rails any either.  If
> GvR wants to improve Python's status against Ruby, I suggest looking at
> what people are *really* raving about in the Ruby world (despite how
> they got there) and address those issues rather than getting sidetracked
> with this nonsense.

First of all, I'd take the raving from the Ruby scene with a pinch of
salt, given the tendency of the blog personalities doing the raving to
breathlessly declare some kind of victory at every turn - as Steve
Holden once said, these people are good at the "don't mention the
weaknesses" style of marketing, and that's probably something various
Python Web framework developers have picked up quite effectively. I'd
rather the Python core developers stopped chasing shadows and looked at
the Python distribution in its entirety. Hopefully, the Python 3000
exercise will see its focus shift into really removing the artifacts of
legacy decisions in both the language and the library rather than
shoehorning more wishlist items into the language.

Paul

P.S. Wishlist shoehorning example: decorators plus annotations ("not
really for static typing, honest!") equals...

http://mail.python.org/pipermail/python-3000/2006-August/003034.html

It's possible that the example decorator in the above message
unintentionally communicates the cynical view of decorators, but the
spectre of multi-page "annotations" is still worth worrying about even
if you're a decorator fan. See other messages in that thread for some
of the ideas (and to bring forth that longing for simpler times).