Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!samsung!zaphod.mps.ohio-state.edu!swrinde!ucsd!nosc!cod!neerma
From: neerma@cod.NOSC.MIL (Merle A. Neer)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: Keep-Alive within TCP
Message-ID: <2261@cod.NOSC.MIL>
Date: 18 Sep 90 21:35:03 GMT
References: <1990Sep7.002637.6209@ingres.Ingres.COM> <59344@bbn.BBN.COM>
Reply-To: neerma@cod.nosc.mil.UUCP (Merle A. Neer)
Distribution: na
Organization: Naval Ocean Systems Center, San Diego
Lines: 17

On a local net the lack of keepalives is a problem in the
sense that if a problem exists in the connection 99.9999999% of
the time its because the other side crashed...i.e. has lost
all knowledge of the connection.....the remedy will be that
tcp will reform the connection but in the meantime valuable
resources (particularly on PC-type hosts) are not available
to other potential connections (PC's are strapPed for buffer
space and file handles)...In the larger internet if a connection
has problems the problem might be in a packet switch or phone
line somewhere and yes, here, tcp 's philosophY Of keeping
the connection alive has validity...and the server/clients
might benefit from keeping knowledge of the connection around.
In lieu of working keep alives we have found it necessary
to program a keepalive capabilitY above tcp in order to
keep PC servers operable (due to the above mentioned phenomena).

Merle