Dmitry Gutov <dgutov@yandex.ru> writes:

> Let's measure it anyway, because the potential impact is big (an extra
> delay when opening any file).

Sure.  I've set up a directory structure with a hg repo inside a git
repo (tar file included as an attachment).

Here's the benchmarks on a local machine and a remote machine with the
current code:

(benchmark-run 1000
  (vc-responsible-backend "/tmp/git-dir/dir1/dir2/hg-dir/bar"))
(0.47081299800000004 10 0.07627535899999999)

(benchmark-run 100
  (vc-responsible-backend "/ssh:stories:/tmp/git-dir/dir1/dir2/hg-dir/bar"))
(2.8259669379999997 99 0.912024865)

Benchmarking for the rewritten code to follow, once I've rewritten it.  :-)

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no