Jeff King <p...@peff.net> writes: > Ultimately I think the path forward is to have a ref backend that > behaves uniformly (either because it avoids the filesystem, or because > it encodes around the differences). See: > > http://public-inbox.org/git/xmqqvb4udyf9....@gitster.mtv.corp.google.com/ > > and its reply.
Once Michael's packed-refs backend stabilizes, we may have a nice calm period in the refs subsystem and I expect that this will become a good medium-sized project for a contributor who does not have to be so experienced (but not a complete newbie). It needs to: - add icase-files-backend, preferrably sharing as much code as the existing files-backend, in refs/. - design a mechanism to configure which refs backend to use at runtime; as this has to be done fairly early in the control flow, this will likely to use early configuration mechanism and will probably need to be done in the set-up code, but doing it lazy may even be nicer, as not all subcommands need access to refs. Thanks for a pointer to the archive.
