git.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Jeff King <peff@peff.net>
To: Kaartic Sivaraam <kaartic.sivaraam@gmail.com>
Cc: Git Users <git@vger.kernel.org>
Subject: Re: protocol v2: More data transmitted between client and server since v2.20.0
Date: Mon, 28 Jan 2019 16:23:05 -0500	[thread overview]
Message-ID: <20190128212305.GA23587@sigill.intra.peff.net> (raw)
In-Reply-To: <CA+ARAtpAN_DJ-zgiwPEBqV1EotgsmggRRQWB59u8O_OPR_kFrw@mail.gmail.com>

On Mon, Jan 21, 2019 at 11:45:20AM +0530, Kaartic Sivaraam wrote:

> I recently came across a blog[1] about how protocol v2 speeds up the
> transfer between the client and the server. It states that the amount
> of data transmitted between the client and the server is less when
> using the protocol v2.

Yes, but...

> # Git version: 2.19.0
> $ GIT_TRACE_PACKET=1 git -c protocol.version=2 ls-remote
> git@gitlab.com:gitlab-org/gitlab-ce.git master
> 2>git_protocol_2_git_2.19.0_stderr.txt
> d4d4ebadfb373518013382560b1f505eb6217f13        refs/heads/master
> 
> $ wc -l
> 22 git_protocol_2_git_2.19.0_stderr.txt
> 
> 
> # Git version: 2.20.0
> $ GIT_TRACE_PACKET=1 git -c protocol.version=2 ls-remote
> git@gitlab.com:gitlab-org/gitlab-ce.git master
> 2>git_protocol_2_git_2.20.0_stderr.txt
> d4d4ebadfb373518013382560b1f505eb6217f13        refs/heads/master
> 04845fdeae75ba5de7c93992a5d55663edf647e0
> refs/remotes/remote_mirror_630f74462b3b08a952486da866d5e702/master
> 04845fdeae75ba5de7c93992a5d55663edf647e0
> refs/remotes/remote_mirror_655ad545056a2ad17e7ebc5461a986e4/master
> d4d4ebadfb373518013382560b1f505eb6217f13
> refs/remotes/remote_mirror_d612bbe5bee4fbc624df371bc7caa759/master
> 
> $ wc -l git_protocol_2_git_2.20.0_stderr.txt
> 160971 git_protocol_2_git_2.20.0_stderr.txt

v2.19, etc, were buggy. The rules for ls-remote's pattern matching do
not permit us to use ref prefixes (because the documentation specifies
that it matches "master" at the end of _any_ ref, not just in the usual
spots). But we did anyway. The fix is in 631f0f8c4b (ls-remote: do not
send ref prefixes for patterns, 2018-10-31).

That should explain the extra refs, as well (they all match "master",
too, and it was wrong that v2.19 did not show them).

It's rather unfortunate, since otherwise ls-remote makes for a nice test
of the advertisement code. ;) You can see some difference with
"ls-remote --heads", as explained in 6a139cdd74 (ls-remote: pass
heads/tags prefixes to transport, 2018-10-31).

The best test is to do a noop git-fetch. E.g.:

  # just make sure we have all the objects. You still see the benefit
  # without it, but if you really want to count bytes, it makes sure
  # you're comparing apples to apples.
  git fetch

  # this has a big useless ref advertisement
  GIT_TRACE_PACKET=$PWD/v0 git -c protocol.version=0 fetch origin master

  # and this is much smaller. Unfortunately it's not just a single line
  # (for "master"), because tag-following requires that the other side
  # tell us about tags, too.
  GIT_TRACE_PACKET=$PWD/v2 git -c protocol.version=2 fetch origin master

  # this one is really tiny
  GIT_TRACE_PACKET=$PWD/v2-notags git -c protocol.version=2 fetch --no-tags origin master

With my origin as https://github.com/gitster/git (which has more heads
than the regular git.git), I get:

  $ wc v[02]* | head -n -1
  2441  17098 340884 v0
   733   5828 130263 v2
    22    141   2037 v2-notags

-Peff

  reply	other threads:[~2019-01-28 21:23 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-01-21  6:15 protocol v2: More data transmitted between client and server since v2.20.0 Kaartic Sivaraam
2019-01-28 21:23 ` Jeff King [this message]
2019-02-02 18:01   ` Kaartic.Sivaraam
  -- strict thread matches above, loose matches on Subject: below --
2019-01-20 21:30 Kaartic Sivaraam
2019-01-20 21:35 ` Kaartic Sivaraam

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20190128212305.GA23587@sigill.intra.peff.net \
    --to=peff@peff.net \
    --cc=git@vger.kernel.org \
    --cc=kaartic.sivaraam@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).