Right after my keynote, Fredrik Lundh was awarded the
Frank Willison Award for his many contributions
to Python and the PYthon community, not just through his
code (e.g. PIL, SRE, xmlrpc, Tkinter and Unicode) but also through his
standard library book and his Daily Python-URL.
Tim O'Reilly's keynote pointed out a new class of "desktop
applications" that run as well on Linux as on Windows: Google, Amazon,
Ebay. Think about it. What these have in common is not just that they
are websites that use open source and dynamic languages to access a
huge database: As Tim points out, their success in a large part comes
from how they track what *people* do. Google's pagerank algorithm
takes into account who links to your website; Amazon and Ebay of
course take direct input, but also reflect indirectly how people use
them (e.g. Amazon's "other books bought together with this book"
feature).
I'd like to connect that directly with Steve Holden's lightning talk
in the Python track: he proposes to build a database of how Python
packages are used together by individual users, with an option to
upload e.g. test results. Something like this could transform the
Python Package Index (PyPI), which currently is a passive list of
packages, but which after augmentation with such information could be
showing things like on which platforms a package builds and installs
successfully, with which Python version and which other packages (and
versions). (Of course a button for users to vote on a package
would be good too, but we can do more than that.)
On the last morning of the conference, Steve also interviewed me,
nicely avoiding the most obvious questions ("what's in Python 2.3?").
I hope he sells the interview to a fancy paper publication like the
New York Times; otherwise, it may end up here on Artima.
Another cool lightning talk was by Chris Mueller of a biochemistry
startup whose name currently escapes me: he demoed a user interface an
analytical chemistry application; this allows chemists to review
results of chemical analyses built on wxPython. The demo was
remarkably responsive, recalculating graphs in real time as the
zooming controls were manipulated, but the killer comment came at the
end: it took two months to create, while a similar app in Java had
taken a year!
And let's not forget the opening lightning talk, which I presented
together with Dan Sugalski: the Great Python Parrot Challenge. Dan
believes that at OSCON 2004, Parrot will be able to execute Python
bytecode faster than CPython can. I don't give him a chance. Dan
bets me ten bucks, a round of drinks, and a pie at ten paces.
My two favorite Python talks at OSCON this year were Adele Goldberg's
and Dennis Allison's "2002: A Zope Odyssey" and Kirby Urner's "Python
in Education". Kirby impressed and entertained by showing how simple
it is for example to introduce polynomials and operations on them in
Python (thanks to operator overloading and polymorphism) and how to
make this into a highly useful and motivating math lesson by adding a
touch of 3D graphics. Adele (yes, her of Smalltalk fame) and Dennis
tackled the other end of education by combining Python, Zope, Flash,
and some Smalltalk (of course :-) to build a system for creating and
running high quality on-line instruction in topics such algebra and
calculus. The whole system was programmed in 4-6 months by Adele and
Dennis (and a part-time Flash programmer) without any prior knowledge
of Python or Zope. Pretty impressive, especially since the resulting
system caters to three different groups of users: a team consisting of
master teachers and animators who create the class, the teachers who
teach the class, and the students who take it.
I had hoped to take at least the morning of "Ruby in a day" taught by
Dave Thomas, but thunderstorms on the east coast messed up my
flight schedule and I didn't make it to bed until 3AM -- really 6AM
according to my east coast body clock. But at least I downloaded
the Ruby interpreter and started reading the on-line help (which is
most of the Addison-Wesley Ruby book, minus a few graphs and with
a few bugs due to the conversion to Windows Help). Ruby is cute,
but too Perlish for my taste.
Later that night I got to ask Ruby's creator, Mats, whether Ruby
was more like Python or more like Perl, while Larry Wall was standing
next to us. Mats responded that it was more like Perl, adding that
Larry was his hero. Yet from what I've grasped so far by skimming the
tutorial, much of Ruby, especially its runtime, is much closer to
Python's (with the exception of Ruby's mutable strings, which I find
an abomination). This was at a dinner organized by (surprise!)
Microsoft, with the purpose of bringing together the creators of
dynamic open languages (and a few other key open source technologies)
and the .NET developers. I got the impression though that Microsoft
hadn't sent real developers but rather a set of evangelists, who were
no match for the open source delegation. It was nice enough to
compare the yield statements in Python and C# (they are very similar,
even though C# quotes CLU as the origin rather than Python :-), but
for me the high point of the evening was Miguel de Icaza's excitement
over Python's feature which makes text files iterators yielding the
lines of the file in succession, so that you can write
for line in open(filename):
...process line...
Miguel is on the ECMA standardization committee for C# and IL (as a
guest of IBM), so if you see something like this in the next revision
of C# attributed to, say, COBOL, you know where it really came from!
Miguel doesn't think I have to worry about Dan Sugalski's pie; in
his opinion, Parrot is based on religion rather than facts. Maybe we
can up the ante a bit?
Another cool keynote was the Eclipse presentation. It's high time
that Jython finds its way into Eclipse as a scripting extension. Kevin
Altis, who coined this idea, says that everyone on the Eclipse team he
spoke thought it was a good idea (or at least nobody said they thought
was a bad idea :-), but what's lacking is somebody to take the
lead. Any volunteers?
Some other talks that I enjoyed: Anna Ravenscroft's review of a
number of Python tutorials from the perspective of a professional
trainer who taught herself programming in Python recently (I saw this
at EuroPython); her fiancee Alex Martelli's presentation of the
Template Design Pattern (horrible name, great pattern, he says); Laura
Creighton's review of the PyPy project, brought with her
usual enthousiasm; and Steven Knight's presentation (at a BOF, where
he had to pay $250 for video projection!) of SCons, a build system written in
Python (based upon his earlier attempt, Cons, which was written in
Perl). I couldn't get much in to A-A-P, the competing project by
vim's creator, Bram Molenaar, also implemented in Python. Apart from
the name, which is meaningless except if you're Dutch, the Perlish
syntax of Bram's Makefile equivalents bugs me. SCons' use of pure
Python for the same purpose makes more sense IMO. An interesting
twist was somebody's suggestion to add SCons (or perhaps a subset) to
distutils, so that Python setup.py scripts can describe a much richer
set of files, actions and dependencies.
Previous Topic
