I enthusiastically support this idea. Even if the project fails, it will make an impact, I believe.
I worked as a packager and maintainer for the Fink project for about a year. The major problem with tiers, as I experienced in Fink and as hes been popularized in Debian, is that everything ends up in "testing" or "unstable". It's really hard to get the confidence to push something to "stable", especially if the packager only uses a subset of the functionality him/herself.
By pushing the "stable" decision back upstream *and* letting the author change that decision post-facto, the tiers will be much more meaningful and authoritative.
The proposal seems heavyweight for what PAUSE/CPAN have traditionally been about, but I think plan to simply let the version number drive the stable/unstable decision in the absence of an author tag will make this be opt-in and thus successful.
looks like a very good idea.
one thing that i feel missing in our current CPAN design is common version number convention that should be followed by people for their package/distribution.
for example, i have seen packages with first version as 1.00 or as 0.01 or as 1.0 or as 0.1.
also, i have seen version increment something like,
so users might find difficulty if they are doing things apart from just installing from cpan (like creating pre-built packages from cpan for their use).
so, i guess it would be useful if CPAN sets common version number convention that people should follow for their package/distribution while releasing new version.