Date: Sun, 21 Mar 1999 01:40:01 -0800 (PST) From: "Kevin S. Ho" <foo@gnv.fdt.net> To: freebsd-ports@FreeBSD.org Subject: Re: ports/10682: List mirror sites in MASTER_SITE_BACKUP - ftp.freebsd.org is often full Message-ID: <199903210940.BAA15667@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
The following reply was made to PR ports/10682; it has been noted by GNATS. From: "Kevin S. Ho" <foo@gnv.fdt.net> To: Kris Kennaway <kkennawa@physics.adelaide.edu.au> Cc: foo@fdt.net, freebsd-gnats-submit@FreeBSD.ORG Subject: Re: ports/10682: List mirror sites in MASTER_SITE_BACKUP - ftp.freebsd.org is often full Date: Sun, 21 Mar 1999 04:38:32 -0500 (EST) On Sat, 20 Mar 1999, Kris Kennaway wrote: > On Fri, 19 Mar 1999 foo@fdt.net wrote: > > 1) Listed extra sites (ftpN.freebsd.org) in MASTER_SITES_BACKUP - often > > ftp.freebsd.org is full - minor fix. > > 2) Allows you to use bwtest to sort sites in order of speed, if you've > > got bwtest - otherwise just uses echo (no sorting) > I couldn't find bwtest as a port - could you point me to it? This sounds like > an idea I was going to work on when I got time. Since you've already jumped > in: Oops on my part. I wrote it, all my friends have it, I forgot I never released it. I'll send you the latest version from my work box monday, it's pretty trivial, but i changed a little bit of it. It's a really stripped down ping (complete with bsd copywrite) that sends four packts of different sizes to a host, and uses them to find 1/bandwidth independent of latency, and then takes a sum-of-quares of it to take into account how badly upspikes hurt ftp. > rtt each time you want to compile a port is hugely inefficient, I was thinking > of a once-off file generation event which runs through all of the mirrors and > orders them according to connectivity. Then instead of a static MASTER_SITE_* > variable, it would work through the top 5 entries in the generated file. Three problems: 1) this won't deal with the large majority of ports who's site is not on a popular, well known archive. 2) This generatiion event will be extremely slow over a modem (if we include all the obscure mirrors of gnu or sunsite or CERT, we're looking at say 400 sites, 5 seconds apiece) 3) Network loads vary greatly. If i run the test in peak time in the US and i'm in the US, i may find optimal being in say germany just because of the time load few solutions seem to exist without an additional layer of indirection ksh To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-ports" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199903210940.BAA15667>