[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Extract of irc log 'the rsync issue'

From: Sander Striker <striker_at_apache.org>
Date: 2001-10-11 20:14:48 CEST

Extract from an old irc log that I have been meaning to send for quite a
while...
But, you know how it is. Time, wish there was more of it.

Sander

[18:14] <striker> Ben, Mike, how does svndiff work (in short)?
[18:16] <cmpilato> it's kind gzippy.
[18:16] <pilchie> svndiff or vdelta
[18:17] <striker> svndif, the stuff sent over the wire.
[18:17] <ben_> careful there
[18:17] <pilchie> I meant which does the compression, vdelta, or svndiff
[18:17] <ben_> svndiff is a way to represent vdelta data
[18:17] <ben_> vdelta is the algorithm
[18:17] <ben_> does binary diffing/compression
[18:18] <ben_> the *output* of vdelta is svndiff data
[18:18] <striker> ah
[18:18] <ben_> and that's what we send over the wire
[18:18] <striker> So everything is vdelta.
[18:18] <ben_> but svndiff data could represent any scheme, really.
[18:18] <ben_> yup
[18:18] <ben_> look in svn_delta.h
[18:18] <ben_> at the top
[18:19] <striker> Will do. Time constraints. While your machine is
building, it is easier to ask one who knows from the top of his head.
[18:19] <ben_> that shows how we create "operations"
[18:19] <ben_> hehe
[18:19] <striker> Sorry about that.
[18:19] <ben_> hey, I built fine
[18:19] <ben_> built shared with new tools, no problem.
[18:19] <ben_> no, I mean I built *static* with new tools, no problem
[18:19] <ben_> cmpilato, however, is building shared, and having probs
[18:19] <ben_> hm
[18:20] <pilchie> i built shared with a little problem in apr which went
away when i updated
[18:20] <pilchie> apr.
[18:20] <striker> Hmm, post 1.0, would it be worthwhile investigating an
rsync like diff (mostly usefull when there isn't a pristine version hidden
in the wc)?
[18:20] <ben_> ah, striker!
[18:21] <ben_> yeah
[18:21] <ben_> if we ever have an option to *not* have a hidden text-base
[18:21] <ben_> that makes sense.
[18:21] <ben_> but I bet that option will never be written. :-)
[18:21] <striker> Ack. :)
[18:21] <pilchie> well i know gregh wants it. but I don't know how badly.
[18:22] <pilchie> It would be useful for universities, where they have lots
of network
[18:22] <striker> If it will, I am happy to donate code and help code it up.
[18:22] <ben_> if you want a clear description of vdelta
[18:22] <ben_> its in libsvn_delta/vdelta.c
[18:22] <pilchie> bandwidth, but small disk quotas
[18:22] <ben_> beautiful comments
[18:22] <ben_> yeah.... I mean, as Karl and I have said privately:
[18:22] <ben_> if anyone actually complains that their working copy is 100
instead of 50 megs
[18:22] <ben_> we'll buy them a new disk.
[18:23] <ben_> disk is just *so* cheap, that it's not worth the trouble
[18:23] * cmpilato doesn't like that his working copy is 100 megs instead of
50. can he have new disk? :-)
[18:23] <ben_> network pipes are the modern bottlenecks
[18:23] <pilchie> Well, as a fourth year Comp Eng student, I get a total of
25MB disk quota
[18:23] <ben_> so it makes sense to optimize around that
[18:23] <ben_> yikes! 25M?
[18:24] <pilchie> Yup!
[18:24] <striker> And there is the justification... :)
[18:24] <striker> I only had 10MB at my uni before I left it.
[18:24] <pilchie> Other engineering depts get around 5
[18:24] <ben_> perhaps karl and I have a skewed view of the world
[18:24] <striker> That includes webspace etc.
[18:24] <ben_> we perceive disk as infintitely cheap
[18:25] <pilchie> Well, a uni is going to pay for a T3, but they don't want
to have massive
[18:25] <pilchie> SAN's to give all users a reasonable amount of space.
[18:25] <ben_> I guess we also assume that most programmers have their own
machines...
[18:25] <striker> Universities are sometimes a bit weird. Ah, like Kevin is
saying :)
[18:25] <ben_> which is a bad assumption
[18:26] <striker> Well in the developers community there are a lot of
students. We might want to keep them in the loop :)
[18:26] <pilchie> well _I_ do, I only use the school machines as xterm's,
but that is not
[18:26] <pilchie> everyone.
[18:26] <striker> One of samba-tngs core developers does everything at uni.
[18:26] <ben_> perhaps someone should file a 'no text-base option' as a
post-1.0 feature
[18:27] <striker> He only has a slow box at home (with expensive dialup).
[18:27] <ben_> hm
[18:27] <striker> Relatively expensive: he's a student.
[18:27] <striker> OTOH he has root privs at the uni, so diskspace isn't his
problem.
[18:28] <ben_> yah.
[18:28] <striker> But in general students having only a machine at the uni,
will have a problem.
[18:28] <pilchie> I _wish_! Yet another case of extremely poor sys-admining
is my school's unix network.
[18:28] <ben_> oh?
[18:28] <pilchie> Don't get me started
[18:29] <striker> Err, Elrond is one of the lucky few. Most people don't
get root privs.
[18:30] <striker> Anyhow, it is best ranked as a post 1.0 issue.
[18:31] <pilchie> As an example of school sysadmin, look at my $PATH.
[18:31] <pilchie> that's about have of it.
[18:31] <pilchie> sorry, try that again:
/home/synopsys/synopsys/syn/sparcOS5/syn/bin:/home/kapilch/bin:/home/ece454/
project/bin:/packages/top:/software/gnu/bin:/software/staroffice/bin:/softwa
re/ghostview-1/bin:/software/ghostscript-4/bin:/software/vim-5.7/bin:/packag
es/cvs/bin:/software/ssh/bin:/software/pine/bin:/.software/local/.admin/bins
/bin:/software/vim/bin:/.software/local/.admin/bins/bin:/.software/local/.ad
min/bins/bin/latex:/bin:/usr/bin:/usr/ucb:/u
[18:31] <striker> sheesh!
[18:32] <pilchie> like i tried to say, that is about 1/2 of it.
[18:32] <pilchie> note all the /.software hidden directories where useful
programs are kept.
[18:33] <ben_> yikes!
[18:33] <striker> Hmm, why hide if you can set permissions?
[18:33] <striker> Lets not go into this, they can't be saved.
[18:33] <ben_> man oh man
[18:34] <striker> Is one of you going to file the issue?
[18:36] <ben_> nah, you do it.
[18:36] <ben_> and make it post-1.0
[18:36] <ben_> note that it's not just a matter of using rsync
[18:36] <ben_> the working copy library is full of comments like
[18:36] <ben_> /* someday, if no text-base exists, use an alternate
strategy here... */
[18:36] <ben_> so it's a big coding job
[18:37] <ben_> but certainly we've left the possibility open.
[18:37] <striker> Noted.
[18:37] <striker> That's why it should be post 1.0.
[18:38] <ben_> :-)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:36:44 2006

This is an archived mail posted to the Subversion Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.