You suggest the bug://domain/local format, but that does not give any indication about what software is affected. The submitter domain isn't as relevant as the software to which the bug applies. As such, bug://apache2/keyword-description might be more useful. Reassigning or retitling would leave forwarders in place (think HTTP 302), and you could do the same for merges. Only clones would be a little more ugly.

I do like the URL scheme and would love to have a bug browser (or just an XML DTD) that could just be used to browse bugs without something like debbugs squeezing it all into more-or-less static HTML.

About metadata, I'd advocate storing metadata separately, such that I could keep my own set of metadata, and you could keep yours, and apache2 could keep theirs.

I cannot help but think that we should look at p2p implementations, like edonkey and bittorrent for inspiration for distributed bug tracking. The added challenge is that the copies are modifiably independently. You still need a central entity, or a continuously maintained mesh with neighbour information to keep it all up-to-date. Now I can't help but think OSPF or BGP.

Anyway, great to see you picking up an interest on this!

-- madduck