On Mon, May 27, 2013 at 1:05 PM, Karen Etheridge <[email protected]> wrote: > I cannot think of any scenario where I would want to block a duplicate > upload from the same user but allow a duplicate upload from another user... > this sounds like a failure in application of the "be liberal in what you > accept" rule.
You're missing what I'm saying. PAUSE prevents a duplicate *path*. It's WORM. You can't upload 'authors/id/E/ET/ETHER/Foo-1.23.tar.gz' twice. You *can* upload 'authors/id/E/ET/ETHER/Foo-1.23-fixed.tar.gz' with the exact same content -- perhaps just fixing your Changes file and having every .pm file be the same. Uploading to 'authors/id/A/AP/APEIRON/Foo-1.23.tar.gz' is just another way of uploading to a different path. > Is this genuinely considered a beneficial feature? Is there any appetite > for tightening up the checking? No, because it would prevent the sort of updates I described. The rule is that module versions must not decrease. There is no rule that requires that at least one version increase. Another example: I take over comaint of Foo. I want to update the Pod to mention that and change the bugtracker. Otherwise, the module is exactly the same. I can ship a new tarball with the same module *code* and thus version. That prevents end users who use cpan-outdated from having to install my tarball just because I changed that. >> No, it's because metacpan is stupid. See >> http://www.nntp.perl.org/group/perl.cpan.workers/2013/03/msg1047.html > > It looks as if metacpan is just following PAUSE's lead here. No. PAUSE does nothing to suggest that distribution names (or name-version) are unique keys[1]. Metacpan, search.cpan, CPAN testers (except Metabase) and RT do. If your DBA came to you with a schema with a unique key for non-unique data (and didn't bother to check or enforce uniqueness), wouldn't you describe that as "stupid" or something similar? "Hi, here's the customer table. I've created an index on FIRST_NAME. We can use it as a primary key for our URL schema and API. Sometimes it's not unique though. Is that OK? It probably doesn't happen very often." /me facepalm David [1] But we addressed this in the Lancaster Consensus and eventually PAUSE will start treating distribution names as unique keys, with some grandfathering of existing distributions. -- David Golden <[email protected]> Take back your inbox! → http://www.bunchmail.com/ Twitter/IRC: @xdg
