also sprach Junio C Hamano <gitster@pobox.com> [2009.02.13.0735 +0100]:
> Once you start making each case arm do more things, it might make sense to
> rewrite the above unrolled loop into something like this:
[...]
> 	} ref_kind[] = {
>         	{ REF_LOCAL_BRANCH, "refs/heads/", 11 },
>         	{ REF_REMOTE_BRANCH, "refs/remotes/", 13 },
> 	};
[...]

> Then we can later add new elements more easily, e.g.
> 
>                 { REF_TOPGIT_BASE, "refs/top-base/", 14 },

As soon as TopGit is integrated into Git proper, this could make
sense. However, I don't know when this will happen. In the mean
time, hardcoding extensions like you suggest might not scale too
well. Wouldn't it make more sense to provide an interface that
allowed tools to register their own namespaces, and handle those
appropriately within Git?

Much of that handling could be taken straight from refs/remotes/*,
as they are conceptually the same. refs/remotes/* just has
additional treatment inside Git, because it's part of the basic
feature set. An external feature's namespace wouldn't be, but Git
also doesn't need to know anything about those, or treat them
specially.

-- 
martin | http://madduck.net/ | http://two.sentenc.es/
 
"doesn't he know who i think i am?"
                                                     -- phil collins
 
spamtraps: madduck.bogus@madduck.net