[Python-Dev] Summaries for Number 2008, First Half

Calvin Spealman ironfroggy at gmail.com
Sun Nov 16 07:39:11 CET 2008

Wow, how long has this been? OK, so these aren't really the summaries.
It is half-done summaries, which I've posted to the wiki for community
collaboration. Anyone with a better understanding a conversation or
simply the knowledge that I don't know what I'm talking about half the
time, the summary can be edited easily now. I'd like to try a schedule
of posting the draft to the wiki, leaving it up for edits for two
weeks, and publishing the final summaries when the next summary draft
is prepared and put up for review. Does that sound like a decent plan
that might actually see the return of the summaries be stuck with?

The draft is at

Of course, if you don't have time to edit this but do want to make any
comments, suggestions, or corrections, or you want to suggest that any
thread should be promoted from the bottom list or demoted down to it,
just let me know. And the current draft is found below.


This is the draft of the python-dev mailing list summaries for the
first half of November 2008. Please make any corrections,
improvements, or comments. The contents will be finalized at the end
of the month, when the summaries' draft for the second half of
November 2008 are posted. I hope to keep this schedule regularly.


Looking for VCS usage scenarios

Brett Cannon solicits any use cases and scenarios that would give an
understanding for the motivations in proposing Python move to or
mirror officially via some distributed version control. Much
discussion ensued that I didn't have the time to follow. This thread
was over 100 posts, so anyone who followed it well, please give as
concise an overview as you can.

Contributing thread:

- `Looking for VCS usage scenarios

Optionally using GMP to implement long if available

The suggestion of using the GMP library to implement Python's long
numbers was brought in the context of an compile time option, falling
back to the current in-house implementation when GMP is not available.
However, the discussion points out that most distribution of Python is
binary and so the LGPL issues that keep us from using GMP still apply.

Some of the conversation also suggestion various improvements to our
own implementation, mostly focused on increasing size of the base long
values are stored in, particularly on 64bit systems.

Contributing thread:

- `Optionally using GMP to implement long if available

Fwd: Removal of GIL through refcounting removal.

Yet Another Attempt At Removing The GIL was proposed, and had a more
positive reception than most of the times it has been brought up (in
my memory). While the removal of the GIL has been both proposed and
successfully attempted in the past, there have always been outstanding
issues that keep the lock around. One of the primary issues is that of
the C API and extension modules, which all expect reference counting.
A technique used in IronClad, a project that provides use of CPython
extensions in the IronPython implementation of the language, was
suggested to provide reference counting semantics to legacy
extensions, even if the internal objects no longer require them.

Contributing thread:

- `Fwd: Removal of GIL through refcounting removal.

Using Cython for standard library?

Without much expectation that any action will actually be taken any
time soon, the possibility of using Cython (and, by extension, similar
solutions) was given consideration. While the benefits of these
solutions are obvious, there is no particular best of breed nor yet a
point of stability that a mature project might stand on. It would be
exciting to one day see some variation of this happen, but that day is
not today.

Contributing thread:

- `Using Cython for standard library?

n.numbits: method or property?

The basic question of a proposed method or property, numbits, drew
into a small discussion about the inherent assumption of binary
representation of our numbers. The numbits method or property would
provide the smallest number of bits required to represent a given

Contributing thread:

- `n.numbits: method or property?

XXX do we need a new policy?

Some concerns were raised for the number of XXX, TODO, and other types
of comment markers in the Python source. While the OP appears to worry
about the number of open issues this represents, the general concensus
seems to conclude that any real issues have associated tickets, there
is little benefit in dealing with all of these at once, and the 2000
or so of them represents nothing significant, in the first place.

Contributing thread:

- `XXX do we need a new policy?

DVCS PEP update

It was announced that git is being included as one of the distributed
version control solutions in consideration for any proposal to move
from Subversion. Some back and forth on the various questions against
each option and who is representing each was also given.

Contributing thread:

- `DVCS PEP update

Released fixes for CVE-2008-2315 for Python 2.4?

It was confirmed that an additional bugfix release of 2.4 will be
made, 2.4.6 in a simultaneous release with 2.5.3. However, this
accepts only security fixes so the multiprocessing module backport is
not eligable.

Contributing thread:

- `Released fixes for CVE-2008-2315 for Python 2.4?

hg branch gone?

Contributing thread:

- `hg branch gone?

Optimize Python long integers

Several outstanding patches to optimize int and especially long
objects were benchmarked and applied to the 2.6 branch for comparison.
Although they are not eligable for the bugfix releases of 2.6 or 3.0,
they may be likely improvements we'll see in 2.7 and 3.1 releases in
the future. However, some related and minor bug fixes to long are
likely to be applied much sooner.

Contributing thread:

- `Optimize Python long integers

file open in python interpreter

Contributing thread:

- `file open in python interpreter

My patches

Contributing thread:

- `My patches <http://mail.python.org/pipermail/python-dev/2008-November/083255.html>`__

compiler optimizations: collecting ideas

Contributing thread:

- `compiler optimizations: collecting ideas

Python 2.5.3: call for patches

Contributing thread:

- `Python 2.5.3: call for patches

datetime and timedelta enhancement

Contributing thread:

- `datetime and timedelta enhancement

Deferred Threads

Previous Summaries

Skipped Threads

- `Why don't range and xrange threat floats as floats?
- `[Python-3000] RELEASED Python 3.0rc2
- `Python 3.0rc2: problem with exec()ing files
- `Feedback from numerical/math community on PEP 225
- `buffer function
- `How to select text of text field in python&#8207;Card
- `RELEASED Python 3.0rc2
- `Summary of Python tracker Issues
- `Getting Set Up dev doc
- `A statistic for Python tickets
- `How does one build Python25.chm on Windows?
- `Upgrade SVN server to 1.5.4
- `[ANN] VPython 0.1
- `Packaging the PyPI version of the SSL module for Debian
- `New "stage" field in the tracker
- `Python2.5 _sre deepcopy regression?
- `test - please ignore
- `AST-level type inference optimizations
- `test message - spam work...
- `Getting Set Up doc now online

Read my blog! I depend on your acceptance of my opinion! I am interesting!
Follow me if you're into that sort of thing: http://www.twitter.com/ironfroggy

More information about the Python-Dev mailing list