Re: TCP alwayskeepalive option.
Vojtech Pavlik (vojtech@suse.cz)
Thu, 9 Dec 1999 22:07:33 +0100
On Thu, Dec 09, 1999 at 12:32:11PM -0700, Richard Gooch wrote:
> TCP keepalives are nasty for machines that suspend. Imagine you have a
> X server running on a machine, and at night you want to suspend the
> machine. If your X clients (running on another machine, say your group
> server) aare well-behaved, they will not generate any traffic, since
> you won't be sending events. So your machine can safely suspend to RAM
> or suspend to disc, and next morning when you resume the machine,
> everything is just as you left it.
>
> Now, with keepalives, the OS on the server will send packets to your X
> server (and expect a response), which of course won't respond. Boom!
> You've lost your remote X clients.
>
> No, I think we want to avoid keepalives. Better to fix Netscape so
> that it exits when the connection to the X server is lost.
On the other hand, TCP keepalives would be great for masquerade - since
now masquerade only relies on timeout. This way idle masqueraded TCP
connections wouldn't die ...
--
Vojtech Pavlik
SuSE Labs
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/