Date: Sat, 23 Mar 1996 07:43:55 +0800 From: Peter Wemm <peter@jhome.DIALix.COM> To: John Polstra <jdp@polstra.com> Cc: jkh@time.cdrom.com, freebsd-hubs@freebsd.org, jkh@freebsd.org, hackers@freebsd.org, current@freebsd.org Subject: Re: new sup server Message-ID: <199603222343.HAA07395@jhome.DIALix.COM> In-Reply-To: Your message of "Fri, 22 Mar 1996 13:37:17 PST." <199603222137.NAA26670@austin.polstra.com>
next in thread | previous in thread | raw e-mail | index | archive | help
>Jordan wrote: > >> > I suspect that Jordan will be trimming off sup's from non-official-mirror >> > sites real soon now to reduce the load on freefall and the T1. >> >> You suspect correctly. > >PLEASE don't do that yet! I am telling you again: the sup mirrors still >do NOT work right. Am I the only one that thinks that using sup to feed the sup mirrors is *extemely* prone to error? The script that is run on freefall takes anywhere between 1.5 and 3 hours to run (depending on what other cron jobs are running in parallel). That's only a 5 hour window of relative safety for the sup mirrors to try and get a clean copy. And, if something goes wrong and a file is damaged on the mirror but the timestamp is not touched, then it stays damaged. And if something happens like the supscan being run from a different directory, it's a disaster. (eg: freefall's sup having src/ prefixes and a mirror doing the supscan inside the src directory with no prefix.) CTM would be *far* better to feed the mirrors with, if only it preserved the timestamps of the files. (phk: hint hint! :-) (note: I'm only suggesting the possibility of geting the data to the mirrors via ctm, not suggesting that the mirrors stop providing sup access once they've got the data.. big difference..) I initially got my source via sup, but after several sup botches (way back at about the 2.0.5 era), I tried CTM. and have *never* had the slightest hiccup in about a thousand deltas. CTM can be a bit of a pain if you're using it on a live source tree, but for mirrors, having the md5 checksums and the sanity checks that go along with it is essential for mirroring. Unfortunately, the lack of timestamp preservation pretty much rules out ctm if it's for a mirrored sup-server which depends on the timestamps. Cheers, -Peter
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199603222343.HAA07395>