[Python-Dev] eggs now mandatory for pypi?

P.J. Eby pje at telecommunity.com
Tue Oct 6 23:19:51 CEST 2009

At 10:17 PM 10/6/2009 +0200, M.-A. Lemburg wrote:
>P.J. Eby wrote:
> > At 06:03 PM 10/6/2009 +0200, M.-A. Lemburg wrote:
> >> It goes a bit in the direction of what we had in mind with writing
> >> for our clients: a tool that looks at the Python installation and
> >> automatically finds/downloads/installs the right package from
> >> our website. Only that we wanted to put up a single index file
> >> for that tool to use, not one per package.
> >
> > That'd get a bit big for PyPI, though, with thousands of listed
> > packages.  It's already pretty slow whenever somebody miscapitalizes or
> > mis-punctuates a package name and easy_install has to read the "file"
> > that just lists all the package names.  That, of course, is more a
> > function of the fact that PyPI's "simple" API doesn't support
> > canonicalization, although I believe certain other PyPI-mirroring and
> > package indexing apps do.
>The above was only meant for our eGenix packages, not PyPI.
>For PyPI the natural answer to this is to use the already existing
>PyPI XML-RPC interface:
>     http://wiki.python.org/moin/PyPiXmlRpc
>This would only have to grow a few more fields for the release
>downloads in order to be able to match the user's Python configuration
>to the available release downloads.

Yes, but it's not easily duplicated by somebody who's just got a 
plain ol' web page, nor is it amenable to static page mirroring.  A 
GET-only REST API, on the other hand, is compatible with both. 

More information about the Python-Dev mailing list