Re: [PATCHv3 00/13] the refs-transactions-reflog series

Michael Haggerty [off-list ref] writes:

... an alternate proposal, to convert
expire_reflogs() to take callback functions that decide *what* to
expire, but which itself does the work of acquiring locks, iterating
through the reflog entries, rewriting the file, and overwriting the old
file with the new one. The goal is to move this function into refs.c and
make builtin/reflog.c much simpler--it will only have to implement the
callbacks that determine the expiration policy.

It sounds like a very sensible approach to me.