I'm working on sketches and notes and angry muttering on the next iteration of CPAN::Mini. I want to keep it fairly simple, but add enough sophistication that it's easier to subclass and plug in. Here are a few things that will be happening:

There will be a "Source" class that will represent the remote CPAN mirror. It will provide a list of modules or distributions and the information needed to retrieve them.

There will be a Mirror or Processor class (or both) that will perform the mirroring from the source, possibly doing extra work via plugins.

There will be (either in the core dist or as an extension) a Source multiplexer that will let you mirror multiple sources.

Output will include a new index (02packages etc) built from the actual mirroring that is performed. Modules you skip will not be listed, for example.

There will be more tests.

I will post more details as they become more than squiggles on my whiteboard.

I also started from same code, and made some changes which might be useful to CPAN::Mini. Take a look at: http://svn.rot13.org/~dpavlin/svnweb/index.cgi/perl/log/trunk/mirror_cpan.pl

I thinks that most useful feature is creation of (optinally compressed) readme file for each module. That allows me to use readme command from cpan shell (which I do a lot :-)

Are you interesed in this change so that I can stop maintaning my own version?

I would like to see the filters redone so that a subclassed filter (which should be a public method) can "rewrite" the entry rather than simply give a thumbs up or down on it. The reason is that I could update CPAN::Mini::Tested to download the most recent version of a module that has passed tests, rather than simply not download a module if the most recent one has not passed a test. (It currently has the side-effect that when a new module is uploaded, it will delete the latest version if the newest one has not yet been tested.)