> 1. Teaching partial clone to attempt to fetch missing objects from > multiple remotes instead of only one. This is useful because you > can have a server that is nearby and cheaper to serve from (some > kind of local cache server) that you make requests to first before > falling back to the canonical source of objects. Quoting the above definition of (1) for reference. I think Jonathan Nieder has covered the relevant points well - I'll just expand on (1). > So much for the current setup. For (1), I believe you are proposing to > still have only one effective <promisor-id>, so it doesn't necessarily > require modifying the extensions.* configuration. Instead, the idea is > that when trying to access an object, we would follow one of a list of > steps: > > 1. First, check the local object store. If it's there, we're done. > 2. Second, try alternates --- maybe the object is in one of those! > 3. Now, try promisor remotes, one at a time, in user-configured order. > > In other words, I think that for (1) all we would need is a new > configuration > > [object] > missingObjectRemote = local-cache-remote > missingObjectRemote = origin > > The semantics would be that when trying to access a promised object, > we attempt to fetch from these remotes one at a time, in the order > specified. We could require that the remote named in > extensions.partialClone be one of the listed remotes, without having > to care where it shows up in the list. Or allow extensions.partialClone=<R> wherein <R> is not in the missingObjectRemote, in which case <R> is tried first, so that we don't have to reject some configurations. > That way, we get the benefit (1) without having to change the > semantics of extensions.partialClone and without having to care about > the order of sections in the config. What do you think? Let's define the promisor remotes of a repository as those in missingObjectRemote or extensions.partialClone (currently, we talk about "the promisor remote" (singular), defined in extensions.partialClone). Overall, this seems like a reasonable idea to me, if we keep the restriction that we can only fetch with filter from a promisor remote. This allows us to extend the definition of a promisor object in a manner consistent to the current definition - to say "a promisor object is one promised by at least one promisor remote" (currently, "a promisor object is promised by the promisor remote"). In the presence of missingObjectRemote, old versions of Git, when lazily fetching, would only know to try the extensions.partialClone remote. But this is safe because existing data wouldn't be clobbered (since we're not using ideas like adding meaning to the contents of the .promisor file). Also, other things like fsck and gc still work.
