I'm glad to see resistance to an off-the-cuff GUID design.
Abstractly, a revision control repository is a collection of revions,
each of which has a name within the repository.
If you give the repository as a whole a unique name, then the two
names combine to give you a global (internet-wide) namesapce.
That, it turns out (c.f. arch) gives you 80% of distributed revision
control.
Nifty. So GUIDs are a good thing.
But look at what they relate too: a global namespace for revisions of
projects; distributed revision control; alternative rev ctl tools
interoperating; bug and issue tracking; distribution inventorying;
peer-to-peer automated testing, .....
There's tons of implications to consider, in other words. arch had to
jump the gate a little here and make up its own kind of GUIDs because
it can't function without them. But even in arch-land, we acknowledge
that one of the reasons we can't call any forthcoming release 1.0 is
because there's a definate need for a bunch of smart folks to sit down
in comfortable chairs in a pleasant room for at least a few days and
talk through as many implications of the structure and interpretation
of global names as we can think of before writing up the formal spec
of their syntax and semantics. It's really not much less serious than
DNS.
Rev ctl is f'ing important. Please help us all take the time to do
it right. Who, aside from ego investment, gives a crap about "svn 1.0
in three months"? And, anyway, isn't it the ultimate ego stroke to
have svn 1.0 that is Clearly Right?
Mission Peak (Fremont, CA) turned green recently. It must be
December,
-t
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Dec 16 13:43:48 2002