httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Justin Erenkrantz <>
Subject Re: [PROPOSAL] Move httpd to the subversion repository
Date Tue, 16 Mar 2004 19:08:40 GMT
--On Tuesday, March 16, 2004 12:26 PM -0600 "William A. Rowe, Jr." 
<> wrote:

> Same here....
> diff -u3 backup/source.c,v live/source.c,v
> mean to say there is an equally trivial way to compare two
> repositories to do post-mortem with svn?  If so please share!

This doesn't work at all in practice because RCS can rewrite the entire file 
if it chooses to making the ,v files non-diffable in any meaningful way.  The 
branches are written out-of-order, and what is full-text once isn't often 
full-text in the next write to the ,v file.  Remember, full-text's are against 
HEAD or the head of the branch-point in RCS.  This increases the S/N ratio of 
the diffs tremendously making them almost useless.

With SVN, it's really a straightforward 'diff -u3 backup.dump live.dump'.

As for doing this with SVN, you'd simply dump your 'live' repository and 
compare the 'backup' dump with the live one.  There should be no differences, 
and this'll work across the entire repository (unlike CVS which asks that you 
do this on a per-file basis).  And, you can't modify any previous revision's 
contents once it is written (unlike editing the RCS file directly) - see below 
for the one caveat to this in SVN.  Any new commits not in the 'backup' would 
be at the end of the 'live' dump file - those can then be visually inspected. 
All of this *greatly* simplifies the auditing process.

None of you are going to be expected to help audit the repository in a 
compromise, but it's us poor fools running the servers that are always stuck 
doing the audits.  And, Subversion would make my life *so* much easier under 
those circumstances if every ASF project were to adopt it.

Oh, as C-Mike pointed out, we'd also have to factor in revision properties 
changing post-mortem - i.e. changing the log message, but we're also capturing 
that data as well, and could easily run a fixup on a backup dump.  -- justin

View raw message