Date: Sat, 5 Jun 1999 20:57:29 -0400 (EDT) From: Garrett Wollman <wollman@khavrinen.lcs.mit.edu> To: Brian Feldman <green@unixhelp.org> Cc: current@FreeBSD.ORG Subject: Re: RE: net.inet.tcp.always_keepalive on as default ? Message-ID: <199906060057.UAA20103@khavrinen.lcs.mit.edu> In-Reply-To: <Pine.BSF.4.10.9906052010540.53878-100000@janus.syracuse.net> References: <199906052344.TAA19843@khavrinen.lcs.mit.edu> <Pine.BSF.4.10.9906052010540.53878-100000@janus.syracuse.net>
next in thread | previous in thread | raw e-mail | index | archive | help
<<On Sat, 5 Jun 1999 20:13:56 -0400 (EDT), Brian Feldman <green@unixhelp.org> said: >> If they are spaced too far apart, it is possible for perfectly >> legitimate connections to get shot down as a result of external >> periodicities. (Does somebody's router reset every day at 2:45? If >> so, better hope no keepalives are scheduled for then!) > But remember that the idea is the keepalive would keep trying for a certain > amount of time, and this would be finely configureable. This wouldn't help the poor sod whose connection gets shot down every eight days while he's not there and doesn't know what hit him. A fundamental principle of protocol design is that synchronization effects can arise totally unexpectedly, with dangerous consequences for the stability of the Internet as a whole. It is necessary to introduce some amount of randomness in order to break up this unintentional synchronization. Modern (post-Van Jacobson) TCP is not directly subject to this effect unless you do something stupid like cause TCP to send a packet like clockwork every X units of time! (TCP can still fall prey to self-synchronization in the applications running on top of it, which is one of the reasons why routers are strongly encouraged to implement a RED queueing discipline by default.) I would withdraw my objection if keepalives were fixed to use a random distribution over [t - t/2, t + t/2]. -GAWollman -- Garrett A. Wollman | O Siem / We are all family / O Siem / We're all the same wollman@lcs.mit.edu | O Siem / The fires of freedom Opinions not those of| Dance in the burning flame MIT, LCS, CRS, or NSA| - Susan Aglukark and Chad Irschick To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199906060057.UAA20103>