On Fri, 14 May 2004, Barbie wrote: > On 14 May 2004 16:59 Randy Kobes wrote: > > > On Fri, 14 May 2004, Barbie wrote: > > > >> On 14 May 2004 15:12 Randy Kobes wrote: > >> > >>> It might be an idea to adjust the automated testing software > >>> so as to ignore testing distributions with names like > >>> PPM-some-other-cpan-distribution.tar.gz. These are ppm > >>> (binary) packages of distributions (as created, eg, by > >>> Module-Build), and thus can't have tests within them. > >> > >> Having 'PPM' at the front of the name doesn't mean it's a > >> PPM package. PPM [1] and PPM-Repositories [2] aren't, as well your > >> PPM-Make [3] :) > >> > >> [1] http://search.cpan.org/dist/PPM/ > >> [2] http://search.cpan.org/dist/PPM-Repositories/ > >> [3] http://search.cpan.org/dist/PPM-Make/ > > > > That's true - my comments applied to an archive that had a name like > > PPM-some-other-cpan-distribution.tar.gz > > If they just contained a blib/ directory then this should > be possible, but the tester would still have to > downloaded/unpacked/check the distribution before that can > be ascertained. > > Perhaps if Module-Build had used something like > '.ppm.tar.gz' (or '.ppm.tgz') as an extension it would > have been easier to spot. Making it look like a regular > tarball is misleading for users as well as automated > software.
That's a good point ... PPM itself doesn't care what the archive name is, as long as it can unpack it. But there may (?) be problems on Windows with an extension other than .tar.gz when using tools to unpack such archives. > > as Module-Build becomes more > > popular, this problem will surface with others, and > > labelling them as failures is misleading. > > Perhaps it would have been wiser to think about the > conflicts such a naming convention would cause. While this > may have been discussed on a Module::Build mailing list, I > haven't seen any thoughts on this previously on other > lists (particularly the module authors list), so I don't > know if there were good reasons for choosing such a name > convention. But anything that starts confusing > any(one|thing) downloading the module is not good. > > I can see it might have been the idea to use the PPM > namespace to indicate only PPM builds, but does this mean > all the regular PPM module distributions are going to > change their name, so they move out of the PPM namespace? I'm not sure the archive name would cause a conflict, other than with the automated testing software. The PAUSE indices don't pick up such distributions (since blib/ directories aren't indexed), so, eg, PPM-NetStumbler-Wap isn't seen within CPAN.pm (and presumably also not within the CPANPLUS shell). And in any case, the distribution name and the module namespace aren't in principle coupled. > If you know of links to any discussion threads, I'd be > interested to read them. I'd also be intrigued to know if > this means RPMs will start to feature on CPAN. If we are > now breaking the standard messages of installing from > CPAN, then I don't think that that is going to do CPAN any > favours. There wasn't much discussion on the choice of PPM-whatever.tar.gz for Module-Build, apart from choosing a (descriptive) name that's different than the one used for making a distribution intended for installation within the CPAN environment. As for binary packages creeping into CPAN, I remember some time ago the CPAN maintainers worried about that (if nothing else, due to the size of CPAN becoming with such packages), but I'm not sure if that's a concern anymore. The ppm packages are relatively small, and often putting them up on CPAN is a convenient way to distribute such packages, as ActiveState's repository doesn't carry ppm versions of all CPAN distributions, for one reason or another. -- best regards, randy
