On Wed, Mar 26, 2008 at 8:26 PM, Ricardo SIGNES <[EMAIL PROTECTED]> wrote: > I am going to be verbose and pedantic. Please know that I am not trying to > sound like a jerk:
I think I know your style well enough by now (virtually) to know you're not being a jerk. ;-) > In the CPAN, there are dists, which are largely "any understood archive, but > especially one that contains either META.yml or is laid out with modules and > other stuff we recognize." META.yml helps make clear what a dist is. Dists > have prerequisites. Packages don't. That's why it's important to have a > dist > as a thing. Awesome! Now I see what you're getting at. I like the idea of tying this to META.yml a lot. Several things immediately fall out of that: * tarballs without META.yml become "badly formed" distributions * the "name" parameter in META.yml is the key to indexing distributions. Distributions that use a name that has already been "claimed" by a prior META.yml could be treated as "unauthorized" the same way package namespace collisions work today. That would help disambiguate names * a distribution tarball name should be the concatenation of its "name" and "version" from META.yml followed by an archive suffix. That doesn't exist as a Kwalitee metric, but could/should be added. * the META.yml spec could/should be revised to include a "uploaded_by" item that must be the CPAN author ID of the person who uploaded it. That would allow the physical location in a CPAN repository to be determined pretty exactly just from a META.yml > Now, is this useful? Yes. The addition of dependency chain to the "platonic ideal" of a distribution and the use of META.yml as a marker tie it together. > My disagreement with this should be clear from the above. I think that using > 02packages as a source for distribution indexing is going to be a losing > proposition. >From my comments above, I think we're now in agreement. Verbose and pedantic helped. Thanks! David