Dependency discovery

Enrico Zini enrico at
Sat Apr 2 15:52:24 PDT 2011

On Fri, Apr 01, 2011 at 01:46:40PM -0700, Toshio Kuratomi wrote:

> Oh!  Regarding distromatch -- at PyCon, Allison Randall and I were figuring
> out how to match packages in Fedora, packages in Ubuntu, and packages on
> pypi (upstream python code hosting) together.  We came up with sonmething
> pretty interesting.  tarballs.  The name portion of a tarballs/zip file/etc
> can be used to match a large number of source packages together.  Even for
> packages which generate multiple binary subpackages, this should help narrow
> down which binary packages are not even remotely related despite having
> similar names.

Thanks, I agree it's a good idea. In distromatch I'm already collecting
"source package name" information from all the distributions supported
so far.  "Source package name" just matches the tarball name everywhere
as far as I know.

So far, however, I still don't make use of that information. One
intention is, in some future tweaking iteration, to use it to validate
approximate matches: given a list of approximate matches, one can try to
trim it to only those that come from a source package (or tarball) with
the same name.

[end of the reply to Toshio Kuratomi, start of a little preaching to the
distributions crowd]

However, I think that distromatch is already good enough that it's time
to build some prototype on top of it that addresses some real problem:
for example, we have categories and screenshots in Debian that it's
really high time others started to steal.

Those are very, very low hanging fruits now: I'd like to see people make
that little bit of effort and just grab them.

I confess that maintaining distromatch as a tool "per se" does not
motivate me: I would like further tweaking of the matching algorithms to
be driven by the need of perfecting its applications.

Until distromatch gets really used, of course, if anybody else is
intrigued by the idea of tweaking distromatch for the pure sake of
tweaking distromatch, they are more than welcome to do so.

I also wouldn't mind to be able to say "in distromatch, *we* do this",
instead of "in distromatch, *I* do this": I am staying next to an
*extremely frequent* bus route[1], after all :P



GPG key: 4096R/E7AD5568 2009-05-08 Enrico Zini <enrico at>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 490 bytes
Desc: Digital signature
URL: <>

More information about the Distributions mailing list