python-dev summary, Nov 1-15
Wed, 15 Nov 2000 22:55:12 -0500
Python-dev summary, November 1-15, 2000
To comment on material in this python-dev summary, you can simply post
to comp.lang.python / <firstname.lastname@example.org>. These summaries are
archived at <URL:http://www.amk.ca/python/dev/>.
The move to Digital Creations has broken the dam, burying python-dev
in a flood of new ideas. This two week period was very busy,
beginning with a call for feature proposals for Python 2.1.
This summary will therefore be more telegraphic than usual; it would
be too time-consuming to summarize the major threads, which were all
quite lengthy, and the final result would still distort the arguments.
So, if one of the topics is of interest to you, please refer to the
python-dev archives for all the details.
Python 2.1 tasks
GvR listed the PEPs he wants to consider for Python 2.1, and requested
more input: "If you have an idea for a PEP that you think should be
implemented in Python 2.1, or if you want to revive a PEP that's
currently listed in one of the "unattainable" categories, now's the
time to make a plea!"
Various people followed up stating what they'd like to work on: Moshe
Zadka wants to tackle the question of integer division (see below),
AMK wants to use the Distutils to build Python itself, and so forth.
Discussion of most of the topics covered in this summary was triggered
by this call for suggestions. After being closely focused on the release of
Python 2.0 for so long, python-dev is beginning to direct its gaze
toward the distant horizon of the future.
Moshe drafted PEP 228, "Reworking Python's Numeric Model". This PEP
proposes a numeric model for Python that isn't based on C's model,
which ultimately derives from machine representations of integers and
floats. "While coercion rules will remain for add-on types and classes, the
built in type system will have exactly one Python type -- a
The ensuing discussion was scattered, mostly revolving around whether
and when to use rational numbers, and whether floating point literals
should be considered exact or inexact numbers. It's not obvious that
this PEP can be solidified in time for its results to be incorporated
in Python 2.1.
Stackless Python, and microthreads
Some sort of resolution to Stackless Python seems likely for 2.1.
Guido is inclined to take a solution for 90% of the problems: "I still
think that the current Stackless implementation is too complex, and
that continuations aren't worth the insanity they seem to require (or
cause :-), but that microthreads and coroutines *are* worth having and
that something not completely unlike Stackless will be one day the way
to get there..." He then went on to post a strawman API for
Christian Tismer agreed with him that continuations aren't really
necessary. "I'm happy to toss continuations for core Python, if we can
find the right building blocks for coro/gen/uthreads. I think Guido
comes quite near this, already."
And so did Tim: "I don't know of any comprehensible application of
continuations that can't be done without them."
Christian Tismer enumerated the changes to ceval.c made by Stackless:
Finally, Gordon McMillan put up a Stackless intro and tutorial:
The seam that runs between classes, defined in Python code, and types,
defined in C code, has been in Python from the beginning. Various
schemes for bridging the gap were suggested.
M.A. Lemburg suggested adding an extra indirection, letting C types
identify themselves as subclasses of another C type:
Greg Ewing, inspired by Ruby's implementation, suggested adding a
global dictionary to add extra attributes to type instances:
Python's two-level scoping rules have been the subject of debate for a
while. Many new users expect Python to have Pascal-like static
scoping and are surprised when variable references in lambdas or other
nested functions behave unexpectedly. Previously this would have
presented technical problems, since nesting scopes would create
cyclical references, which couldn't be handled by Python's reference
Jeremy wrote PEP 227, "Statically Nested Scopes", to make a concrete
proposal: "The current language definition defines exactly three
namespaces that are used to resolve names -- the local, global, and
built-in namespaces. The addition of nested scopes would allow
resolution of unbound local names in enclosing functions' namespaces."
Some discussion ensued, mostly about whether it's worth changing this
just to fix nested functions. A side thread on dynamic scoping, as
opposed to static, spun off at one point, but since practically
everyone thought it was a bad idea, nothing emerged from it worth
Another item on GvR's list for 2.1 was a framework for printing
Coincidentally, Paul Prescod was also working on a warning framework,
and had a draft PEP:
AMK tried to raise some interest in PEP 222, which lists some
improvements to make Web programming in Python easier, but there seems
to be little interest. If no further interest is shown by the
community, the PEP will simply be abandoned.
A Web page listing critical patches for 2.0 was created:
Python project page on SourceForge:
Python Enhancement Proposals (PEPs):