python-dev summary: Oct. 17-31
Wed, 1 Nov 2000 21:55:55 -0500
Python-dev summary, October 17-31, 2000
(To comment on material in this python-dev summary, you can
simply post to comp.lang.python / <firstname.lastname@example.org>.)
Pythoneers Move to Digital Creations
The most significant news in this two-week span was the five
Pythoneers (GvR, Fred Drake, Jeremy Hylton, Tim Peters, Barry Warsaw)
leaving BeOpen and being hired by Digital Creations.
The terms of their hiring seem quite liberal. From GvR's
announcement: "We will be spending part of our time on core Python
development (including Jython and Mailman) and part of our time on
Python infrastructure improvements that also benefit Zope." And, a
few paragraphs later: "All future work we do on Python as Digital
Creations employees will be owned by a non-profit organization yet to
be created. We think of this new organization as the Python Software
Foundation. In the meantime (while the PSF is under construction) I
will own such copyrights personally."
No official statement from DC has been made as of this writing, though
Paul Everitt has made some informal statements: "I'd like to start by
saying that Guido and I have known each other a long time now. Barry
too. We all really like each other, but we also know how to disagree
There was only a slight amount of paranoia triggered by the news.
Some people wondered if DC would begin dictating Python's development,
or start slanting the standard library toward Zope (discouraging
enhancements to the Web development tools, for example). This
scenario seems quite improbable. Instead, my crystal ball says it's
more likely that certain language features and proposals will be given
more of GvR's attention than they've previously gotten. For example,
Jim Fulton, Zope's primary architect, has long been interested in ways
to specify interfaces in Python:
So interfaces may suddenly become a bit more interesting to Guido.
Mending the class/type dichotomy would also be useful for Zope, and
that means the odds of ExtensionClass or some equivalent
functionality, being added to Python have probably just increased.
(See "Other Matters", below, for an example. ) But I do expect Guido
will keep calling the shots as to exactly how any such new features
fit into the language.
It was an eventful two weeks for scripting languages in general,
really. On October 20th that a company called Interwoven announced
that it will be acquiring Ajuba Solutions, formerly known as
Scriptics. Scriptics was the company founded by John Ousterhout to
provide support for the Tcl scripting language and the Tk GUI toolkit.
However, Interwoven isn't interested in Tcl/Tk at all and will not be
funding its development.
Ajuba knew this was coming, of course, and over the past few months
efforts have been made to open up Tcl/Tk development to the community,
taking such steps as moving the CVS trees to SourceForge, electing a
Tcl Core Team with 14 members, and specifying processes for proposing
and voting on core changes (the Tcl equivalent of PEPs).
However, this process is still in its infancy, and the announcement
caused much worrying in the Tcl/Tk user community about the language's
future and its continued maintenance. This affects Python because Tk,
through the Tkinter module, is the most GUI most commonly used with
Python. Tkinter is included in the source distribution, it's the most
extensively documented, and it's the most portable, supported on the
Big Three platforms, namely Windows, Unix, and MacOS.
Fredrik Lundh first posted a note about the announcement, and Eric
S. Raymond raised the question of what this meant for Tkinter: "This
raises anew a question I've been meaning to bring up for the last
week: is it finally time to move away from Python's dependence on
Tcl/Tk for GUI support?"
People were split on this question, though Guido was receptive to the
idea ("Yes, it may be time to have this discussion again.") and
pointed out some reasons to stick with Tkinter: "Tk has two very high
quality widgets: the Canvas and Text widgets are unsurpassed in
functionality and robustness by other widget sets. You can draw more
lines or characters per second in most toolkits, but few toolkits
offer the convenience of marks and tags, not having to deal with
refresh events, being able to group objects, move them around, change
attributes, etc., etc., etc."
Greg Wilson's reaction was "Yes please", and he went on to explain
what factors kept him using Tkinter for a recent course:
/F thought Tk was still a good choice, and said " ... and trust me, the
Python 2.0/Tk8.4/Tkinter3000 combo will rock ;-)". Tkinter3000
is an improved Tk interface that /F has been working on, with
more flexibility and better performance.
/F also hoped that the Tcl Core Team would become more receptive to
making it easier to use Tk from other languages without Tcl having to
be in the middle, which would let the maintainers of Tkinter and the
Perl/Tk interface participate more in Tk's development.
Eric S. Raymond speculated about the success of Tcl's new development
model: "Good for Tcl: Osterhout's rather lame attempts to develop
under a mixed model have almost certainly come to an end in favor of
an Apache-like model funded by big Tcl users."
Greg Ewing wondered if stdwin should be revived. GvR and Raymond both
thought that far too much work, and not productive of very good
results. "And stdwin would really look pathetic in this time", GvR
Python isn't tied very tightly to Tk, of course; well-maintained and
reasonably complete bindings exist for many different GUI toolkits,
such as Qt, GTk+, wxWindows, Windows MFC, and a few more.
wxWindows is probably the most obvious second candidate, since it
actually supports all of the Big Three platforms, and no other toolkit
does. Robin Dunn, author of the wxPython binding, posted to discuss
the status of wxWindows on the Mac: "My understanding is that the
version in CVS is nearly up to date with the features in the MSW and
GTK versions, though I haven't had a chance to use it myself. ... The
next step I guess is getting it wrapped up in wxPython..."
There was no clear final decision, and my crystal ball didn't deign to
give an answer. Given all the uncertainties, it's probably best to
wait and see. If Tk's development continues to progress and becomes
more open to languages other than Tcl, Tkinter will probably continue
to be the most common Python GUI. If not, we can consider this
question in another 6 months, which will give wxWindows and wxPython
time to develop on the Mac platform. And who knows? Maybe something
else will happen, such as Qt or GTk+ being ported to the Macintosh.
Paul Prescod suggested that we should start thinking about how to make
the transition to the vaporware Python 3000 easier: "Rather than
delaying painful backwards-compatibility breakage I think we should
make it less painful. We should decide where we are going long before
we ask our users to move there. I think we should start including
alternatives to syntaxes and features that we know are broken. Once
people move to the alternatives we can 'reassign' or remove the
original syntax with much less pain."
Reaction was enthusiastic, GvR commenting: "I think the proper approach is to start a separate migration process
for each of these proposed changes. Each should be paced separately
(the pace may depend on how hard to swallow the change is for users as
well as how hard it is to implement the new functionality) and for
each, a separate PEP in the 3000 series should be started."
Paul got approval to write PEP #5: "Guidelines for Language Evolution".
Along the way, Tim Peters commented: "Not just to be my usual self
<wink>, but I do see a from-scratch rewrite as being less likely as
the years go by. There's nothing I know of in Guido's plans that
can't be done incrementally instead -- and if he doesn't either,
selling a total-rewrite project to an employer is probably impossible.
The most successful projects I've seen and been on *did* rewrite all
the code routinely, but one subsystem at a time."
Neil Schemenauer posted a proposal for a core change that would make
ExtensionClass better able to simulate a Python instance:
"Extension classes can never pass the PyInstance_Check predicate.
I've searched for all occurances of PyInstance_Check in the 2.0
source. In many places that PyInstance_Check occurs a more
general 'is this an instance-like type' check would seem to be
Guido thought this simply papers over the problem in this one
particular case, and said "Rather than formalizing the exceptions made
for instances, we should work on eradicating the differences between
classes and types. I believe that the plans for rich comparisons
would or could also take care of this. I hope I'll have more time
soon to explore this." Jim Fulton, in a brief followup, said "Me
Barry Warsaw proposed adding a function to the sys module which would
return the stack frame for getting the stack frame of the function's
People liked the idea; M.-A. Lemburg, Greg Ward, and Ka-Ping Yee all
had their own implementations of this function already. ?!ng reminded
the list of his inspect.py module which provides this function, and
other functions for inspecting live Python code, and suggested it for
inclusion in the standard library:
No resolution was arrived at.
Python project page on SourceForge:
Python Enhancement Proposals (PEPs):
What's New in Python 2.0: