Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 14 Mar 2001 15:19:30 -0700 (MST)
From:      Nate Williams <nate@yogotech.com>
To:        Dan Nelson <dnelson@emsphone.com>
Cc:        Dag-Erling Smorgrav <des@ofug.org>, freebsd-arch@FreeBSD.ORG, Nate Williams <nate@yogotech.com>
Subject:   Re: [PATCH] add a SITE MD5 command to ftpd
Message-ID:  <15023.61042.768406.854325@nomad.yogotech.com>
In-Reply-To: <20010314161646.A1482@dan.emsphone.com>
References:  <20010314105918.A5204@roaming.cacheboy.net> <35525.984597779@critter> <20010314210758.A2405@roaming.cacheboy.net> <15023.53743.215996.538067@nomad.yogotech.com> <20010314130025.A3031@dragon.nuxi.com> <xzpelw0yqun.fsf@flood.ping.uio.no> <20010314161646.A1482@dan.emsphone.com>

next in thread | previous in thread | raw e-mail | index | archive | help
> > > The checker can *easily* keep a list of files sizes and date stamps
> > > and compare that.
> > 
> > Date stamps are useless. They'd be different on different master
> > sites anyway. File size is almost as bad, as two files can very
> > easily be totally different and still have the same size.
> 
> But how often do port distfiles change, but keep their size?

Often enough that it's been seen.

> Pretty low, I'd say, at least compared to the number of times the size
> changes and the filename stays the same.

True, but now we'd have to modify every port to include both the MD5
*AND* filesize information in the port.  Right now the MD5 is part of
the port, so all the information is necessary in the ports tree to do
the SITE-MD5, except for the ftp infrastructure.

 
> Another thing to consider before adding SITE MD5 as a command is that
> it's an extremely slow operation.  md5'ing a 10MB file takes about
> 1/3rd of a second on my pIII/600.   It would take 5 minutes of CPU time
> to md5 1-gig worth of sources, and that's assuming that the FTP server
> is idle.  ftpd would have to cache the md5 checksum somewhere for it
> to be acceptable, and then you've got the same caching problem (how
> does ftpd know when the file has changed to is can update its cached
> md5?).

Is that cost greater than the cost of sending the data out over the
wire?



Nate

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-arch" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?15023.61042.768406.854325>