[Catalog-sig] PyPI down?

"Martin v. Löwis" martin at v.loewis.de
Tue Jan 12 23:20:45 CET 2010

>>> Is it being considered as the starting point, or not?
>> Not by me, no. I think meta-data mirroring is completely pointless
>> in order to achieve PEP 381's objective.
> I don't quite understand that statement; the meta-data is part of the
> mirror.

Only if you apply a textbook definition of the word "mirror". For what
PEP 381 calls mirroring, you don't need the meta-data at all.

> If z3c.pypimirror is not the starting point then is there a different
> implementation that will be used as a 'jumpstart'?

I'll write it from scratch, using urllib (or perhaps Twisted). We
evalutated z3c.pypimirror at the last PyCon, and I came to the
conclusion that it is not appropriate as a starting point: it does
too little (not mirroring everything that PEP 381 tells it to), and
it does too much (also grabbing packages/distributions/parcels
from other places - i.e. not from PyPI), to support offline operation
(which is not the objective of PEP 381).

> I have read it.  I'm mirroring the meta-data in order to make an easy
> way for tools that want to pull the meta-data to do it without
> beating the crap out of the xmlrpc interface on pypi.

But you don't need that. All you need is the list of project names
to start with, and then the only XML-RPC operation will be to look
at the changelog. You don't need any of the other package metadata
(I don't quite recall why "pypi" is listed as a URL to mirror; IMO,
it shouldn't).


More information about the Catalog-SIG mailing list