Why is bzr pull via http soooooo sloooooow?

[Michael Ellerman]

On 6/13/06, Stefan Reichör <stefan at xsteve.at> wrote:
> Hi!
>
> I monitor the development of bzr. Today I run bzr missing & bzr pull
> to get an up to date copy:
>
>   # time bzr missing
>   Using last location: http://bazaar-vcs.org/bzr/bzr.dev/
>   You are missing 14 revision(s):
>   ------------------------------------------------------------
>   revno: 1765
>   committer: Canonical.com Patch Queue Manager<pqm at pqm.ubuntu.com>
>   branch nick: +trunk
>   timestamp: Mon 2006-06-12 02:12:14 +0100
>   message:
>     Stop using shlex in run_bzr_subprocess
>   ------------------------------------------------------------
>
>   snip...
>
>   ----------------------------------------------------------
>
>   revno: 1752
>   committer: Canonical.com Patch Queue Manager<pqm at pqm.ubuntu.com>
>   branch nick: +trunk
>   timestamp: Thu 2006-06-08 07:38:32 +0100
>   message:
>     (jam,mbp,wvh) Lots of updates to unicode,url,and encoding support
>   bzr missing  0.75s user 0.09s system 0% cpu 4:00.91 total
>
>
>   # time bzr pull
>   Using saved location: http://bazaar-vcs.org/bzr/bzr.dev/
>   All changes applied successfully.
>   14 revision(s) pulled.
>   bzr pull  4.89s user 0.54s system 0% cpu 21:03.38 total
>
>
> So missing took 4 minutes and pull took 21 minutes for 14 revisions!!
> I used a rather recent version (bzr (bazaar-ng) 0.9.0dev0).
>
> So my questions are now:
> * Is bzr really that slow?
> * Am I doing something wrong?
> * Do I need some extra libraries to speed things up?

Yes, probably, maybe. ;)

No, it's currently pretty slow, especially over high latency links.
Hopefully in the next few days I'll have some code ready to speed up
http a bit. Keep an eye on the list if you're interested in testing
it.

cheers