Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!tut.cis.ohio-state.edu!ucbvax!SUMEX-AIM.STANFORD.EDU!mrc From: mrc@SUMEX-AIM.STANFORD.EDU (Mark Crispin) Newsgroups: comp.protocols.tcp-ip Subject: (response to message of Tue, 30 May 89 11:59:13 EDT) Message-ID: Date: 4 Jun 89 20:09:39 GMT References: <8905301559.AA29126@monk.proteon.com> Sender: daemon@ucbvax.BERKELEY.EDU Organization: The Internet Lines: 14 I think the pro-keepalive people were expressly stating that an ICMP error should not shoot down a connection. I hate Unix's behavior in doing this -- it kills several perfectly good connections/day on me every working day -- and I patched my Unix workstation not to do it. But it doesn't help if the guy I am talking to is still doing it. All we want is a mechanism to tickle a TCP reset iff the other end has already punted the connection. In a sense, this is not a keepalive as much as it's a clean up if already dead. TCP should do this. And, by doing this, it may prevent people from implementing the kinds of keepalive and timeout mechanisms that we *all* hate. -------