Date: Wed, 14 Mar 2001 23:42:36 +0000 (GMT) From: Terry Lambert <tlambert@primenet.com> To: roam@orbitel.bg (Peter Pentchev) Cc: tlambert@primenet.com (Terry Lambert), freebsd-arch@FreeBSD.ORG Subject: Re: [PATCH] add a SITE MD5 command to ftpd Message-ID: <200103142342.QAA09233@usr08.primenet.com> In-Reply-To: <20010314084651.A23104@ringworld.oblivion.bg> from "Peter Pentchev" at Mar 14, 2001 08:46:51 AM
next in thread | previous in thread | raw e-mail | index | archive | help
[ ... avoid downloading wrong file when name has not changed ... ] I think it would be better to put metadata in the archive, at the front, and just downloa tthat, and abort the transfer if it's the wrong one. I now understand that this is intended to handle that very special case, and only that case, even though people will probably depend on it for "security" anyway, if it's there. > However, if this were to be implemented, on however few of the sites, > then it *could* provide a great opportunity even for the maintainers > themselves - it is very easy to imagine a remote FTP MD5 checksum > tool, that is run by each port maintainer on a regular basis for each > of his ports, so things do not get to the point that annoyed users > blame the maintainers, the Ports Collection and the FreeBSD project > for the deeds of dumb authors. I'm of the opinion that if you were a site, and you supported a large number of connections, it would not be in your best interests to implement this feature: it has dubious value at best, and it costs you resources to do the calculation. Put another way, for 5,000 clients, I'd rather have each one run the checksum locally, rather than me having to run it 5,000 times, only to have it accepted and downloaded anyway. I think this would end up making my FTP server CPU bound instead of I/O bound. It's already memory bound, so it would steal some memory from me, which I could otherwise use to support additional client machines. Also, since FTP transfers are linear without reuse, reading once for the sum and again for the transfer is going to thrash my buffer cache, costing me even more CPU and latency to thrash them back in. If you had to have this, it would be much better, since large FTP servers are generally write-mostly, to build the MD5 into the FS metadata, and calculate it each time a file is closed after having been modified/created. This would let the ftpd do a low cost fcntl()/stat() operation to get the information, instead of making it run an expensive calculation. I think it's like GNUtella: if a few people have it, it's good; if everone has it, it falls apart under load. Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers. 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?200103142342.QAA09233>