Re: [PATCH v6 03/19] index-helper: new daemon for caching index and related stuff

David Turner [off-list ref] writes:

Per working tree/repository, yes (not quite per-index because working
trees with split indexes have two).

Heh, I consider "split index" merely an implementation detail of a
(logical) single index file that happens to be represented with two
physical files.

Anyway, thanks for clarification.

Why do we try to avoid select(2)?  It seems marginally less efficient
for small numbers of fds, and might not scale to large numbers of fds,
but we're not using it in a performance-critical way, and we only care
about one fd at a time.

Just that having to deal with two APIs that do very similar things
is less preferrable than just using one that was designed later by
those who learned lessons while using the older one.