[darcs-devel] adding feature: patch origins

[Stephen J. Turnbull]

Ganesh Sittampalam writes:

 > One thought is that it's a bit at odds with the distributed nature of
 > darcs to store this information only in a specific repository; I wonder
 > if it's actually the state of a repository per se that's important here,
 > but rather the state of one particular branch. For example the official
 > version of a piece of software or the official state of the development
 > branch or whatever.

Those don't exist in Darcs as currently implemented, only in the minds
of Darcs users.  For example, does the order of application of
commutative patches matter?  Many Bazaar users will tell you it does,
and in fact they often claim that such history should be written in
stone.  Git users agree that it does, but they tend to think of it as
a matter of presentation (ie, rebasing), and therefore essentially
editable.  Darcs doesn't care by design.  So, I don't know what the
state you're talking about "really is," but if you call it "branch"
or "version" you confuse (and, I suspect, offend :-) users of DAG-
oriented VCSes.

Experience shows that when such things matter, most organizations and
individual users are perfectly happy to identify the branch (in W3C
terms, an URN) with a "canonical" repository (do., URL).  (Those who
aren't happy use git. ;-)  Similarly, for particular versions, users
identify the version (URN) with a tag in the canonical repository (URL).

 > I think these issues could possibly be addressed by storing the
 > tags off to one side instead of in the actual repository, but the
 > details would need thinking through.

I don't understand what "actual repository" is supposed to mean here,
given that there's a suggestion that this should be reflected in
clones etc.