[dist-bugs] Use cases, etc

[Pierre Habouzit]

On Wed, Jul 09, 2008 at 08:42:34PM +0000, Joey Hess wrote:
> Aidan Van Dyk wrote:
> > All the below looks great in fantastic bright colours, but what about
> > the transient nature of the web and all it's URLs?
> 
> Obviously it's a weak point. The best I've been able to come up with in
> this scheme is that:
> 
> a) The aggregators archive *everything*. Content of every url they see.
>    (Possibly legal issues here.)
> b) The scheme does allow multiple aggregators to serve as copies of a bug
>    report, so if the urls to one go down, another can take over.
> 
> If the aggregator does go down, urls pointing to it do of course break,
> and this could leave a non-buginfo-aware part of things (like the web
> forum or gitweb in my example) with a dead url, and no way to get to the
> new url (other than web searches).
> 
> Actually, I don't care if forums, mailing list archives, etc, end up
> with dead links -- urls left in such things will, by their nature, rot
> over time, and its no worse than links from forums to bug urls now.
> 
> In my example, the git repo did have a bug tracker living in it, that
> did get a copy of all the bug status[1]. So while having old urls to
> bugs in git commit messages sorta sucks (but is no worse than including
> urls in commit messages for any reason today), if you look up the same
> bug in the tracker you'll still get good data. Including, presumably,
> the new urls for the bug.
> 
> So the best thing would be to have your commit message include both a
> url for the bug, as well as the ID for the bug in your local tracker.

  This is what GIT solved using recomputable sha1's for their objects.
Once an object is in git, it's the same UUID for everyone. Moreover, if
you think of an URL like a UID and not like a pointer anymore, one could
have what you call aggregators that are "official" mirrors for that URL
and remember about those for life.

  Once you have that, you have no "URL" pointing to them, but rather
something like: dvcs://some-url.used-as-an-id.com/<bug-uuid> and you
would have some kind of well known name server that maps the fqdn to
real hosting. You can also declare that sources.redhat.com is a full
alias to sourceware.org or things like that easily, it would be fully
transparent, and resilient enough.

  Of course bug trackers that were never mirrored will go down. But if
you care so little about your bug tracker that you let it break, then
the software it's used for is probably worthless :P


-- 
·O·  Pierre Habouzit
··O                                                madcoder at debian.org
OOO                                                http://www.madism.org
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: not available
URL: <http://kitenet.net/pipermail/dist-bugs/attachments/20080710/16fa8478/attachment.pgp>