[Catalog-sig] Re: [Distutils] PEP 241 draft

Andrew Kuchling akuchlin@mems-exchange.org
Thu Mar 15 12:03:05 2001


On Wed, Mar 14, 2001 at 10:45:25PM -0700, Sean Reifschneider wrote:
>Alternate wording:
>   Developers may not provide their own "METADATA" file.  The "sdist"

More fascist; I like it!  Will add it...

>Questions: Is it expected this name corresponds either to the name of the
>package which is imported, or the package top-level directory name?  If so,

I don't think so; consider "Sketch", which may not actually
have a Sketch package.  

>what do we do about alternative packages that provide different implementations
>of the same functionality?  I suppose we could reasonably expect to make
>use of "import urllib2 as urllib" to take care of that.


>Does this include platform name, platform version, and architecture?  Like
>redhat-7.0-x86, windows-nt-hppa, etc?

I wish I knew!  This is the last remaining XXX in the PEP.  Thoughts?

>   Freely-Redistributable
>   Group (such as "Database", "Network/SMTP", etc)?
>   Provides (maybe "urllib2" would provide "urllib"?)
>   Requires (dependences -- RPM for example has multiple lines of the
>         form "Requires: initscripts >= 3.25", "Requires: openssl >= 0.9.4")

I'm lukewarm; can we decide on syntax for all of those fields?
(Semantics can wait; as long as users can put 'Group: blah', we can
figure out what the groups should be later.)  Can
Freely-Redistributable be derived from the License field?  (Side
effect: discourages using your own Open Source license).  Will
comma-separated values be sufficient for the last 3?

--amk