[ocaml-biz] the Python competitor

[Brandon J. Van Every]

Brian Hurt wrote:
> Brandon Van Every wrote:
> >
> > The question with Python is whether anyone's going to solve its
> > performance problems over the next 3 years.  People are
> > working on it.
>
> This gives us our window closing, then.  Ocaml needs to get
> popular before
> Python gets fast.  Having a window open now doesn't mean
> it'll stay open.
>
> What are the approaches people are taking to make Python
> fast?

The main one is a thing called Psyco.  http://psyco.sourceforge.net/  It
speeds up Python quite a bit, with no annotation to the Python program
at all.  However, its speedups aren't dramatic enough to put it in the
ballpark of C++.  If they were, I would have never moved on to OCaml.

There's also something called Parrot.  Recently the author of the Parrot
virtual machine challenged Guido to a benchmarking contest.  The loser
would get a pie in the face at the Python conference, hence Pie-thon.
http://www.hole.fi/jajvirta/weblog/20040108T2001.html
http://www.sidhe.org/~dan/blog/archives/000371.html
Parrot lost.  It beat cpython (the standard implementation from Guido et
al) on 3 of the tests, but it failed to complete a significant number of
tests.  Which would seem to say, Python could definitely perform better
than the standard cpython implementation, but in the real world, you
have to get that done without breaking everything.

Something to be scared of is Guido Van Rossum and Larry Wall have
written a book together about it.  http://www.oreilly.com/parrot/

> A python
> front end to GCC would give python Ocaml-level performance at little
> developer cost, ignoring the inevitable licensing flamewars.
> The window could close in a few months.

I wouldn't pronounce such gloom and doom.  I'm not up on the innards of
Python, what can or can't be done with it, but I doubt the problem is as
easy as that.

Also, remember these clowns throw business leaders and web developers
away.  They can't market their way out of a paper bag, and I doubt
they've yet learned from their mistakes.  If we can get on with decent
looking logos and websites, and we don't tick off INRIA, we'll have
accomplished a lot more in terms of marketing infrastructure than Python
has accomplished.

> A much harder case to make, considering that Python's scale
> up (once it's
> performance problems are solved), while I don't think would
> be as good as
> Ocaml's, won't be the absolute train wreck like some other languages
> (Perl).

Someone else has got a static type checking project for Python.  I can't
find a link to it right now though.

Also, the need for strong typing is highly debateable.  An entirely
different school of thought is that unit testing is what actually
matters.
http://www.xoltar.org/misc/static_typing_eckel.html
Python people can and will make this argument, successfully.

I don't think we need to actually spend any time yakking about "we're
better than Python."  Python is just this inexorable clock ticking.  It
has a critical mass, it has a whole lot of bodies banging away at its
problems.  It has trouble with various problems, being what it is, but
eventually those problems are going to yield to sheer brute force.
Assuming nothing came along to supplant it and capture attention.


Cheers,                         www.indiegamedesign.com
Brandon Van Every               Seattle, WA

"We live in a world of very bright people building
crappy software with total shit for tools and process."
                                - Ed Mckenzie