[Catalog-sig] PEP 314: latest draft

[Mark W. Alexander]

On Tue, 29 Apr 2003, Andrew Kuchling wrote:

> Here's the current draft of PEP 314.  A number of XXX comments have
> been removed; version declarations can now be included in the
> Conflicts and Obsoletes fields; the Classifiers field from PEP 301 
> is now included.
> 
> An open issue: with the addition of Classifiers, should the 
> Platforms and License fields be deprecated and/or removed?  

Someone already mentioned that License should be required, so I agree
that it should be retained independent of the Classifiers.

Platforms could go either way, so let me throw out another idea:
Platforms could be a some kind of attribute that maps the package name
to the differing naming requirements of binary packagers. For example,
Debian packages must be all lower case. Solaris packages have (an
extremely annoying, and I know you agree with the annoyances of
Solaris ;) 9 character limitation. This also implies that "linux2" is
insufficient as a "platform". Something that indicates the OS native
package format would be more helpful.

It's my firm belief, however, that the metadata should contain all the
information that's required for those who would like to write bdist
commands, and with the exception of the previous paragraphs annoyance,
I think you've covered that.

Now another question for thought. The metadata corresponds to what?
The source distribution or a binary package or both? I'd like to see
Distutils progress to where multiple binary packages can be produced
from a single source, ala Marc's mx series, only without the need to
create custom setup classes (Marc's far more creative than I am). If
we can ever get there, then the metadata would be most usefull if it
was available for each binary package. Again, _if_ that's a reasonable
target, then source distributions will need to be able to provide
multiple instances of the metadata, one for each potential binary
package.

mwa
--
Mark W. Alexander
slash@dotnetslash.net