Path: utzoo!utgpu!water!watmath!clyde!cbosgd!ucbvax!PARK-STREET.BBN.COM!brescia
From: brescia@PARK-STREET.BBN.COM (Mike Brescia)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: Hop problem
Message-ID: <8801250356.AA00959@ucbvax.Berkeley.EDU>
Date: 25 Jan 88 03:24:52 GMT
References: <314777.880122.JBVB@AI.AI.MIT.EDU>
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 21

The Internet Protocol Time to Live (TTL) field was introduced as a way of
detecting and supressing loops.  Ideally, a packet would be timed as it passes
from source host through gateways, packet switches, satellite links and to the
destination, and further within the destination while awaiting fragments for
reassembly in the IP layer.  Since clocks are not yet universally syncronized,
the only thing we see being done is decreasing the TTL within each gateway.
The IP spec. states that a gateway must decrease this field by at least 1
(otherwise it could possibly live forever).

Now, when the TTL is viewed as a field decremented by each gateway a packet
passes, it appears to be a hop count field.  Packet radios or packet switches
are actual hops in a network, but do not alter the TTL field, even though we
sometimes saw queueing delays in the Arpanet in the 10-20 second range.

      packet-radio people are talking about ... a 30+ hop routing

I think Phil Karn said something about that path being only about 10 seconds
long.  Also, I expect that the radio hops were not doing gateway functions, so
they did not count the IP TTL field.

mike