I've just tried out latest darcs from the unstable and converted our repo to hashed format: http://darcs.ramaze.net/ramaze.hashed

For some reason, get --lazy via ssh takes an hour, while the same from http takes 7 minutes. A complete `scp -r` is in between. Wouldn't it be possible to improve speed of get, maybe by using a remote command to tar.(bz2|gz) the whole repo and d/l that or using scp -r?

I tried following:
scp -r ramaze.net:ramaze.hashed .  1.21s user 8.50s system 0% cpu 38:51.73 total
darcs get --lazy ramaze.net:ramaze.hashed ramaze  6.08s user 4.97s system 0% cpu 59:59.34 total
darcs get --lazy http://darcs.ramaze.net/ramaze.hashed ramaze.hashed.lazy.http  1.09s user 2.41s system 0% cpu 7:33.57 total

On Sat, Feb 16, 2008 at 04:31:39PM -0000, Michael Fellinger wrote:
> I've just tried out latest darcs from the unstable and converted our repo
> to hashed format: http://darcs.ramaze.net/ramaze.hashed
> 
> For some reason, get --lazy via ssh takes an hour, while the same from
> http takes 7 minutes. A complete `scp -r` is in between. Wouldn't it be
> possible to improve speed of get, maybe by using a remote command to
> tar.(bz2|gz) the whole repo and d/l that or using scp -r?
> 
> I tried following:
> scp -r ramaze.net:ramaze.hashed .  1.21s user 8.50s system 0% cpu 38:51.73 total
> darcs get --lazy ramaze.net:ramaze.hashed ramaze  6.08s user 4.97s system 0% cpu 59:59.34 total
> darcs get --lazy http://darcs.ramaze.net/ramaze.hashed ramaze.hashed.lazy.http  1.09s user 2.41s system 0% cpu 7:33.57 total

I believe we talked on IRC about this a bit, and you had installed darcs2
on the remote server, but had trouble getting the transfer-mode connection
to set up (you posted a transcript from darcs get --lazy --debug).  I've
added some better debug output, so would you mind trying the darcs get
--lazy --debug again and post the transcript? It looked like the pipe
wasn't getting set up properly, and I'd like to track down why.

I've added a number extra debug messages when things fail, in particular
for when the transfer-mode is not working, so hopefully this will let us
track down the problem.

Thanks for your bug report, by the way!
-- 
David Roundy
Department of Physics
Oregon State University

I did as you asked, pulled latest darcs-unstable and tried again:
http://p.ramaze.net/584
I cannot seem to find the relevant parts though.

Assigning to David so that he can look at the log that manveru helpfully produced.

Hi Michael, are you still experiencing the problem of darcs get over ssh being
very slow?

No word from Michael since February.  Marking presumed-dead (but please do feel
free to re-open).  Thanks!

Whoops! I just realised that my last contact was on 20 August.  Sorry, Michael,
re-opening.

I've taken a look at this log and don't see any clues about what happened with
darcs transfer mode.  It just says that we tried to invoke it, and then moved on
to using scp a lot.

I know David has recently fixed http://bugs.darcs.net/issue1003 but it does not
seem to be related as far as I can tell (the repository being fetched is hashed).

Michael: it would be good if you could test this with darcs 2.1.0 on both the
server and client side, but since we still haven't heard from you in a while, I
am going to mark this presumed-dead (whilst still hoping you can get back in
touch with us).  Thanks!