[Catalog-sig] RFC: PEP 243: Module Repository Upload Mechanism

Sean Reifschneider jafo@tummy.com
Tue, 20 Mar 2001 17:46:37 -0700

On Tue, Mar 20, 2001 at 04:13:08PM -0800, Amos Latteier wrote:
>First off, I think it may be too early to draft this PEP. It may make
>more sense to wait until we have one or more catalog prototypes

Well, I *DID* demo my prototype catalog server at Python 9...  The problem
I face is that it's not really very useful until we actually have stuff
*IN* it.

I also see the upload facility is really orthogonal to the issues of the
actual catalog.  You don't need a catalog server to implement the upload,
and therefore I don't see any real reason for delaying this for the catalog
server.  Having an upload facility in the 2.1 distutils will allow us to
get some real-world data for the catalog server.

>PEP (rather than evolved as we work with catalog servers) then we should
>spell out some requirements and use cases.

How is that different from what has been going on with the discussions
around my swalow project over about the last month?  It's been on the
distutils, catalog, and python mailing lists...

>Was the name of the catalog server decided at the conference? Did we
>decide who's in charge of it, etc? 

The name "modules.python.org" was suggested by AMK, I believe.  At the
current time, it probably makes sense to set it up on one of my boxes, and
I'll take responsibility for it...

>For source packages this can be extracted from the archive.

Yeah, and I actually have the code written to do this for .tar.* files.  I
wasn't sure I wanted to require that the upload server implement that
though -- I'll take this as a vote that it should.

>>         platform (optional) -- A string representing the target
>>         platform for this distribution.  This is only for binary
>>         distributions.  It is encoded as
>>         "<os_name>-<os_version>-<platform architecture>".
>Why not just use the platform meta-data? I see no need for two different
>and incompatible ways to specify platform information.

Sorry, what platform meta-data are you speaking of?

>I'd like to talk to you about the catalog server. I have built several
>prototypes myself, and I'd love to talk with you more about the project.

I thought that's what you were doing.  At the moment I have a working
prototype of a catalog server using the working name "swalow", which is
available from the ftp.tummy.com FTP site.  The package upload code I also
have prototyped, but haven't yet released it.

 "I'll thrash you like a Netscape process on a machine with 640K."
                 -- John Shipman, 1998
Sean Reifschneider, Inimitably Superfluous <jafo@tummy.com>
tummy.com - Linux Consulting since 1995. Qmail, KRUD, Firewalls, Python