On Jun 2, 2012, at 9:24 PM, Marvin Humphrey wrote: >> You know I wrote something like this, right? >> >> http://pgxn.org/ > > For sure! And since you are a fellow Apache Lucy developer, you may have > recognized the system I referenced as Clownfish[1].
No. I though Clownfish was a thing for integrating C code with a dynamic language, not a module distribution platform. > PGXN and Clownfish are both off-topic for cpan-testers-discuss, except that > perhaps in this case it might be nice to hear about any experiences with PGXN > and distro-vs-sub-distro dependency specification that would help us to > understand the design tradeoffs better, and ultimately to develop best > practice guidelines for CPAN authors and users. Got anything for us? PGXN allows extensions to have version independent of the distribution version. Even those most distributions so far have only one extension, many already have variance between distribution and extension versions: http://pgxn.org/dist/twitter_fdw http://pgxn.org/dist/semver/ The reason for this is that often an individual extension does not change at all, so it's not worth it to bump its version, because, for PostgreSQL extension support, that would mean that people would be forced to upgrade for no reason (much more of a deal for a persistent database than for an in-memory language). It also means that authors would have to write upgrade scripts that do nothing and include them in the distribution (and their repos). Easier to just to allow there to be a divergence. So I think it depends on the platform. I personally strongly prefer to have the version in all modules match the distribution version in my Perl modules. But obviously there are disadvantages to that for PostgreSQL extensions. Maybe there would be for your thing, too. Or maybe not, since I am ignorant of your plans so far. Best, David
