To: gnu-arch-users@gnu.org
From: Charles Duffy <cduffy@spamcop.net>
Date: Mon, 06 Dec 2004 14:18:07 -0600
Message-ID: <pan.2004.12.06.20.18.06.379046@spamcop.net>
References: <20041205000613.PNEI7152.lakermmtao09.cox.net@nonerjsnum1tkq>
	<26E1F314-4654-11D9-AD55-000A957659CC@spy.net>
	<20041205023828.GA11443@suffields.me.uk>
	<87pt1o9kvb.fsf@tleepslib.sk.tsukuba.ac.jp>
	<200412061850.iB6IoWmS036469@xl2.seyza.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 8bit
User-Agent: Pan/0.14.2.91 (As She Crawled Across the Table (Debian GNU/Linux))
Subject: [Gnu-arch-users] xdelta/rdiff/friends (was Re: Arch Versus CVS
	Versus Subversoin)
Precedence: list
Sender: gnu-arch-users-bounces+migo=homemail.com@gnu.org
Errors-To: gnu-arch-users-bounces+migo=homemail.com@gnu.org
Status: RO
Content-Length: 1101
Lines: 14
In practice, I've found that xdelta1 works surprisingly well in most
binary-diff situations; likewise its neighbor rdiff (from librsync; rdiff
allows rsync's diff-generation algorithm to be used locally). xdelta3 has
some bugs which make it generate severely suboptimal output (over a vastly
lengthened execution time) on hitting certain corner cases, which makes it
IMHO unusable even if its performance and memory-usage characteristics are
better than those of xdelta1 (which I find in practice are still more than
adequate for any uses I've tried to put it to).

rdiff and xdelta1, though, are in my experience surprisingly effective in
terms of the set of cases where they can generate a usable delta. Were the
other associated problems to go away, and the feature become enough of a
priority to justify the needed development time, I'd certainly be
comfortable with their use.


_______________________________________________
Gnu-arch-users mailing list
Gnu-arch-users@gnu.org
http://lists.gnu.org/mailman/listinfo/gnu-arch-users

GNU arch home page:
http://savannah.gnu.org/projects/gnu-arch/