Stackless Python, eventual merge?

[Martin v. Löwis]

Ian Bicking <ianb at colorstudy.com> writes:

> Was there some indication that Stackless would be integrated if the code
> was cleaned up, documented, or otherwise simplified?  I don't believe
> there was, but I may be wrong.  

I certainly meant it that way in

http://groups.google.de/groups?q=g:thl2323460570d&dq=&hl=de&lr=&ie=UTF-8&oe=UTF-8&selm=j4d72xasx0.fsf%40informatik.hu-berlin.de

> Without such an indication, there is no reason to think that mere
> technical changes to Stackless will lead to it being included in
> standard Python.

This is not how free software works. Most contributors don't ask for
an advance blanket incorporation guarantee, and nobody would ever give
one, for any kind of change. Instead, they just contribute the code
they find useful, and then are perhaps told to modify it. If they want
to make sure that the feature itself is undebated, they need to write
a PEP.

There are many features in Python that most maintainers won't care
about (for example, I have no interest in IMAP). That hasn't stopped
people to contribute those features, and hasn't stopped maintainers to
incorporate them.

Contributing needs patience, a skill that contributors sometimes
lack. To give a few examples:

- the IPv6 patches took about 15 months to integrate. For many months,
  the patches were unreviewed, because nobody thought to have the
  expertise to review them. I eventually tought myself the necessary
  background to review the patch, after which about 4 further
  iterations where necessary before the patch was installed, and about
  10 changes after the patch was installed. 

- To build Python as a shared library on Unix, about 4 patches have
  been contributed over several years. Each contributor was told what
  the requirements for such a patch were (must be a compile-time
  option, must work uniformly on many systems, etc), most contributors
  abandoned their patch after learning these requirements. Only the
  most recent such patch could be incorporated, after a few
  iterations.

Both patches were relatively small, compared to the old Stackless
implementation.

Since integrating Stackless will require a significant amount of time
on the side of the Stackless authors, some contributor must step
forward and be willing to guide the entire integration period,
preferably for two Python releases after the patch has been
incorporated. I don't think anybody was willing to contribute that
much time in the past.

Regards,
Martin