Subject: Re: svn disk usage: where we are and where we're going
To: Martin Husemann <martin@duskware.de>
From: Eric Gillespie <epg@NetBSD.org>
List: tech-repository
Date: 09/14/2007 10:20:51
Martin Husemann <martin@duskware.de> writes:

> FWIW, I understood the problems raised here different: svn
> (maybe only in older versions) seemed to be a runtime memory hog, so
> slow in everyday use.

The only slowness is from the wc crawls.  On a fast file system
such as ext3, the first one on a really large tree can take up to
a minute, very annoying.  But afterwards, they're all nearly
instantaneous.  Sure, it still sucks, we know.  The wc is broken,
and will be fixed.

The memory issue is an entirely separate one, though also a
result of the wc design flaws.

I was addressing neither of those issues, but only the space
issue.  I've already talked about those two, but now there's been
a lot of confusion on the developers list about disk usage.

> And it used significantly more disk space (as you explained) for every
> checkout - while other systems only used that once (for the replicated
> database).

As others already explained, you can use svk, and only pay for a
(partial or full, your choice) repository replication.

> I completely do not understand how this is related to older hardware 
> or support for other archs in any way.

Your next paragraph is it.

> It is certainly a problem for people working on a lot of branches or having
> multiple work areas checked out simultaneously.

Disk space is cheap on modern platforms.

-- 
Eric Gillespie <*> epg@NetBSD.org