[21:40] <danderson> inserting an extra diff stream would make the
svndiff library go "huh?" when trying to apply diffs of later revisions,
because of new offsets etc.
[21:41] <danderson> It's certainly not impossible to write a tool that
"redeltifies" the whole repository for you
[21:41] <danderson> updating offsets and such in later revs
[21:41] <danderson> but it'd be horribly difficult imho
[21:41] <toad_> but that would cause issues on an update?
[21:41] <danderson> that would at least cause the svn server to tell the
client his working copy is corrupt
[21:42] <toad_> right
[21:42] <danderson> because what the client sees as r100 is not what the
server sees as r100
[21:42] <toad_> cool
[21:42] <toad_> okay, i think we'll go and implement our new repository
now :)
[21:42] <danderson> :-)
[21:42] <nextgens> :)
[21:42] <toad_> thanks danderson 
[21:42] <danderson> toad_: in that case, Welcome to Subversion!
[21:43] <toad_> :)
[21:42] <danderson> toad_: in that case, Welcome to Subversion!
[21:43] <toad_> :)
[21:43] <nextgens> toad_: how many commits do you think we had on CVS ?
[21:44] <toad_> nextgens: about 9000 changesets in the cvs2svn import
iirc
[21:44] <nextgens> I mean how many subversion's rev. would it do ?
[21:44] <danderson> toad_: if you'd like, I'll ping the other devs to
get a more complete explanation of how an attacker could slip in things
unnoticed
[21:44] <danderson> there are several who are uber experts at all the
relevant bits of the code
[21:44] <toad_> danderson: that might be interesting
[21:44] <danderson> (svndiff streams, checksumming and integrity checks)

On Sat, Oct 08, 2005 at 09:40:48PM +0100, Matthew Toseland wrote:
> An interesting point from a friend on IRC:
> [21:36] <danderson> well, first of all, you know that in Subversion
> revision numbers are repository-global
> [21:36] <danderson> so any commit bumps the whole repository rev by 1
> [21:37] <danderson> so even if somehow one mail doesn't go out
> [21:37] <danderson> the next mail to get out will have a rev that is +2
> compared to the last mail
> [21:37] <danderson> instead of +1
> [21:37] <danderson> so you'd detect holes quite fast that way
> [21:38] <danderson> furthermore, you can query the repository for the
> last revision it has
> [21:38] <toad_> okay, and slipping one in would be quite difficult...
> [21:38] <danderson> toad_: try impossible.
> [21:38] <danderson> a missing revnum is just about the biggest integrity
> failure in the repository
> [21:38] <toad_> danderson: why impossible?
> [21:38] <toad_> well yeah but can't he coalesce his change with somebody
> else's?
> [21:39] <danderson> well, that's not insertion, that's alteration
> [21:39] <toad_> ... while that's possible, it would be very obvious to
> somebody checking the changelog
> [21:39] <toad_> cool
> [21:39] <danderson> and he *could*, but he'd need to insert a svndiff
> stream that is both compatible with previous diff streams, AND doesn't
> break the diff streams of followup commits
> [21:40] <toad_> i'll send an email out with part of this conversation,
> suggesting that somebody write such a tool
> [21:40] <danderson> I'm certainly not intimate with the svndiff system,
> but I'd wager it'd be difficult
> 
> On Sat, Oct 08, 2005 at 09:36:14PM +0100, Matthew Toseland wrote:
> > If somebody compromizes the Subversion (or CVS) repository, they can
> > potentially do commits without them going to the commit list, and
> > therefore introduce evil code. Hopefully this will be picked up, but
> > Freenet is quite large. If you want a non-java task to increase
> > freenet's security, I suggest a script that can cross-reference the CVS
> > list emails with the actual log from SVN/CVS, and flags up any
> > discrepancies. If such a thing already exists, I'd be very interested;
> > if it does not, some perl hacker who can't be bothered to learn java
> > could write it.
> -- 
> Matthew J Toseland - [EMAIL PROTECTED]
> Freenet Project Official Codemonkey - http://freenetproject.org/
> ICTHUS - Nothing is impossible. Our Boss says so.



> _______________________________________________
> Tech mailing list
> Tech@freenetproject.org
> http://emu.freenetproject.org/cgi-bin/mailman/listinfo/tech

-- 
Matthew J Toseland - [EMAIL PROTECTED]
Freenet Project Official Codemonkey - http://freenetproject.org/
ICTHUS - Nothing is impossible. Our Boss says so.

Attachment: signature.asc
Description: Digital signature

_______________________________________________
chat mailing list
chat@freenetproject.org
Archived: http://news.gmane.org/gmane.network.freenet.general
Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/chat
Or mailto:[EMAIL PROTECTED]

Reply via email to