[Catalog-sig] simple index and urls exracted from metadata text fields

"Martin v. Löwis" martin at v.loewis.de
Sun Sep 13 15:53:36 CEST 2009


> Take the hachoir-core as an example. The home URL was changed in
> 1.2.1 to :
> 
> "http://bitbucket.org/haypo/hachoir/wiki/hachoir-core"
> 
> the 1.2 version home url was:
> 
> "http://bitbucket.org/haypo/hachoir/wiki/hachoir-core"
> 
> But the PyPI simple API will keep track of both:
> 
> http://pypi.python.org/simple/hachoir-core
> 
> Leading to the problem described (because the script visits all urls
> before it decides
> what tarball to pick)
> 
> So what the maintainer should do ?

One option would be to delete releases that point to versions which no
longer work. Another way would be to make sure that, despite the URLs
not pointing to actual data anymore, they still give a quick meaningful
error message (such as "connection refused"). In the very specific case,
deleting the DNS entry for hachoir.org, or pointing it to, say,
127.0.0.1.

> Recreate a new version of an old release so the old URL is removed
> from PyPI ? Just register the metadata, knowing that the one contained in
> the tarball is not the same ?

The latter would also fix this problem in a convenient way, ISTM.

> Maybe the solution would be to add in that page only the latest home URL link..

That would solve this specific problem. I'm not so sure it would solve
the general problem: what if some package *wishes* to provide different
homepage URLs for different releases, and has them all working
simultaneously?

Regards,
Martin



More information about the Catalog-SIG mailing list