jafo-pythonlist at tummy.com
Tue Dec 18 11:37:49 CET 2001
On Sat, Dec 08, 2001 at 12:11:23AM -0300, Walter Moreira wrote:
>I've been reading some of the catalog-sig archives and I wonder why the
>interest of the community about something like CPAN is so small. When a
The problem is largely related to a lack of critical mass. Nobody uses it
because nobody uses it. This was my opinion in March of 2000 at Python 9
when I showed off my working prototype of a client and server CPAN-like
system, and it would still seem to be the case.
The system I showed back then had the ability to catalog packages based on
their version, platform, and type. For example, it could keep track of
a package's distutils source, i386 binary RPM, HP-PA binary .deb, etc. It
also tracked the idea of a cluster of mirrors.
The client I had would allow you to searcn, download, and/or install the
specified packages. It would repeatedly try random machines in the list of
mirrors automatically, download, build RPMs from distutils source, and
install them. A fairly complete system...
I demonstrated it at Python 9, and then we got together in the cataloging
SIG to speak about packaging further. Active State promised to open up
their tool, which doesn't seem to have happened. I expressed the idea that
I thought packages should, as much as possible, be handled using the native
system packaging format (using RPMs and .debs, for example), and the idea
of needing critical mass. We also talked about making a python-specific
packaging system, which had some benefits and some drawbacks.
On the way home from the conference and the week following it (the week or
so leading up to the 2.1 release), I worked to add code to make the
distutils package upload packages to the server I had set up. It, however,
was never integrated into Distutils...
For example, if you'd normally do:
you could do:
./setup.py --submit sdist
and the source package would be uploaded to my server. AMK had created a
PEP for including basic package information into the dist file, which my
server could read and use for cataloging. About the only thing that my
system didn't include was a security mechanism to prevent the submission of
bogus new versions of existing packages, and testing on a range of
After publicizing the implementation and the "--submit" patches to the the
main python list and the distutils/catalog sigs, I got exactly *ONE*
So, is it that the people who want it aren't the people who can make it
happen? Are the people who want it to happen not motivated enough to
actually do something about it? Is it just not really that interesting of
I wish I had an answer there.
So, currently the system I built is gathering dust. I've set up resources
for Suschandra to host his code. I hope it goes better.
So, if you want to see it happen, I'd have these words of advice: Get
Involved. Get some packages uploaded, use it, advocate it...
Examine what is said, not who speaks. (Arabian Proverb)
Sean Reifschneider, Inimitably Superfluous <jafo at tummy.com>
tummy.com - Linux Consulting since 1995. Qmail, KRUD, Firewalls, Python
More information about the Python-list