On Dec 29, 2004, at 11:35 PM, Richard Jones wrote:

So, is anyone interested in getting together before PyCon for a sprint to
finalise and/or implement PEPs 314, 243 (I guess we should ask for some PSF
money, though we've obviously missed the most recent grant allocation round)
and possibly 262 (I consider this optional and don't believe it should get in
the way of implementing the other two)?


Metadata for Python Software Packages v1.1
  http://python.org/peps/pep-0314.html
Module Repository Upload Mechanism
  http://python.org/peps/pep-0243.html
Database of Installed Python Packages
  http://python.org/peps/pep-0262.html

I'd want to revise PEP 243 at some stage. I want to *only* implement source
distribution uploading for now. Also, there's some aspects of the protocol
that need work (eg. we should use standard HTTP responses, not
X-Swalow-Status).

I'm interested in helping this along. Similar to your revisions with PEP-243, here's what I think should happen to PEP 314:


- It should be required to list every "top-level" package/module that is installed as Provides fields (else distutils will do it for you). "top-level" isn't absolute, as there will be cases like zope.interface. By top-level I mean the highest level that is to be installed by this setup.
- Every Provides must provide their own version (for Requires), distutils will default these to the global version if not explicitly specified.
- Requires should not be arbitrary. They should be exactly Python module names. It just gets too messy otherwise. Having such a meta-requirement is completely and utterly useless without also having PEP-262 implemented and accessible at runtime. If you say that a package requires *some* DBI API 2.0 driver, how would you find one? In order to support this sort of use case, it might be worthwhile to have a way to say Requires "foo (>1.0) OR bar (>1.2)" (say a package that's compatible with PEAK or Twisted, but needs one or the other).


-bob

_______________________________________________
Catalog-sig mailing list
[email protected]
http://mail.python.org/mailman/listinfo/catalog-sig

Reply via email to