Re: please pull ppc64-2.6.git

From: Sergey Vlasov <vsu@altlinux.ru>
Date: Aug 31, 2005, 13:08 UTC
Message-ID: <20050831170852.2f549318.vsu@altlinux.ru>
In-Reply-To: <7v8xyjhxkd.fsf@assigned-by-dhcp.cox.net>

On Tue, 30 Aug 2005 15:25:22 -0700 Junio C Hamano wrote:

Show 16 quoted lines

> Christian Meder <chris@absolutegiganten.org> writes:
> 
> > Is alternates unthinkable with URLs (e.g. remote alternates).
> 
> In order to read an object data, the low level core GIT layer
> does open()/mmap() of a file on the locally mounted filesystem.
> It also does opendir()/readdir()/closedir() to find what's
> available.
> 
> If you can arrange to automount via luserfs (or whatever; we do
> not care much about particular implementation) so that it can do
> these filesystem operations, git will happily use it.
> 
> Even in that case, the alternate you would tell git should look
> like a local filesystem path --- your automounting from remote
> specified with URL for such a path would happen outside GIT.

All this means that currently there is no clean way to publish a partial GIT repository, unless you place it at the same server where the base repository is located (and even in that case needing to use something like "echo /pub/scm/linux/kernel/git/torvalds/linux-2.6/objects > objects/info/alternates" looks like a horrible hack).

What if we create a mapping layer from URLs to local mirrors of corresponding repositories? Store this mapping, e.g., in ~/.git-core/mirrors, then if git reads an URL instead of an absolute local path from objects/info/alternates, it will consult that file and either use the specified local mirror, or complain that a local copy of the base repository is not available. The partial repository will then contain no system-specific paths and could be published without problems.