[BangPypers] Unladen Swallow for Py3k

[Noufal Ibrahim]

On Thu, Jan 7, 2010 at 12:49 PM, Anand Balachandran Pillai <
abpillai at gmail.com> wrote:

> [..]
>
>  What I am more worried is that if Google produces a CPython
>  implementation on *nix which is say 50-100 times faster than
>  the main CPython implementation for most common operations,
>  then with Google's influence and muscle, wouldn't this become
>  the choice of many high end users and not the regular CPython ?
>  That won't bode well for the future of Python as a true "free"
>  language - free as in the sense of 90% of the work not coming
>  from a corporate entity here. Technically it is still free since they
>  have to keep the license compatibility, but implementation-wise,
>  it is Google-Python. In short I am worried about a major "fork"
>  in Python.
>

I personally considered unladen-swallow a fork (although they claim on their
site that it's a branch rather than a fork). Also, given the fact that
they're enhancing a version of Python (2.x) which the official guys are
trying to move people off, it was a problem.

With this announcement though (which as of now looks like vapourware), Py3k
will have performance benefits as well. That would be an extra reason to
move.



>
>  There is not much chance of all these patches getting merged
>  to CPython mainline. First of all I don't think they should approach
>  LLVM as a panacea to fix all Python VM ills - which they seem to be
>  doing.
>

Given the size of project, you're probably right but atleast there seems to
be a nascent plan to do that.



>
>  It would be better if we see an experimental branch of CPython
>  branched from py3k trunk to which many of the more experimental
>  and incompatible bytecode changes go from Unladen Swallow. But
>  as I pointed out earlier, I am a lurker in pydev and have not seen any
>  such discussion yet.
>

This is what Unladen swallow has done but with two differences
0. They branched 2.x rather than 3
1. They have completely separate repo and project site (which makes it like
IronPython or Jython rather than a branch of CPython itself).

The current announcement seems to be a unifying one and that's a good sign.
Whether it will happen is a different question.


-- 
~noufal
http://nibrahim.net.in