Re: [PATCH v2] send-pack.c: add config push.useBitmaps

From: Jeff King <peff@peff.net>
Date: Jun 16, 2022, 21:04 UTC
Message-ID: <Yqua243d7FPqWxQH@coredump.intra.peff.net>
In-Reply-To: <220616.86fsk4ww69.gmgdl@evledraar.gmail.com>

On Thu, Jun 16, 2022 at 03:38:36PM +0200, Ævar Arnfjörð Bjarmason wrote:

Show 8 quoted lines

> But the change as it stands is effectively saying "bitmaps on push
> hinder more than they help, and turning them on for push was a mistake".
> 
> Maybe that's true, but I don't think we've got any data to support
> that. Even though I've got one of those anecdotes (and from occasional
> investigations of slow "git push" I'm pretty sure this would help my
> use-cases more than it would harm them) the plural of anecdote isn't
> data.

Yeah, I'm not convinced they hinder more than they help on net. And any problems on pushing are also things that _could_ happen on fetching, so I think this is really a band-aid. But:

  - even if the average is improved, it's reasonable to want to avoid
    the most pathological cases

  - it's reasonable for it to be the user's choice to make, and right
    now the config isn't expressive enough to allow the choice they want

My biggest fear would be that the pathological cases are improved (perhaps with the approach of mine that you linked earlier; thanks for that), but that people erroneously continue to think that turning off push.useBitmaps is a good idea due to inertia or superstition. But that seems like a minor problem that can be addressed later if need be.

-Peff