phil@BRL.MIL (Phil Dykstra) (03/14/89)
If you run traceroute through the Buttergates in the core, you will notice that they do not decrement the IP TTL. This is not a very good idea. We would be in deep trouble if a routing loop was ever set up between them (not that this would ever happen of course :-). - Phil
brescia@BBN.COM (Mike Brescia) (03/19/89)
If you run traceroute through the Buttergates in the core, you will
notice that they do not decrement the IP TTL. This is not a very
good idea.
Phil Wood raised this in a msg dated 2/10, which appears not to have gotten a
(broadcast) answer. The short of it is that the Butterfly gateways claim to
decrement the TTL, and claim to discard packets received with TTL=0, but do
not claim to avoid sending packets with TTL=0. While they do not yet conform
to RFC1009 in this detail, the Butterflies will not cause a packet to live
forever.
If you want TTL to be a measure of 'how many gateways may touch this packet',
then one with TTL=1 will pass through one gateway. I don't view this as
refusing to comply with RFC1009, but attempting to explain the thinking behind
the current implementation. We made our design decisions before 1009 was
being discussed. In the interests of uniformity, this is the sort of detail
that should be included in protocol definitions.
I hope this explains your traceroute results.
Mike Brescia
Gateway Development Group, BBNCC
------- Forwarded Message
Date: Fri, 10 Feb 89 11:17:35 MST
From: "C. Philip Wood" <cpw%sneezy@lanl.gov>
Message-Id: <8902101817.AA03292@sneezy.lanl.gov>
To: tcp-ip@sri-nic.arpa
Subject: TTL
Why do the "Mail bridges" not decrement the TTL? Or, do they?
(on 26.0.0.90)
#./traceroute ucbarpa.berkeley.edu
traceroute to ucbarpa.berkeley.edu (10.0.0.78), 30 hops max, 40 byte packets
1 UCBARPA.Berkeley.EDU (10.0.0.78) 633 ms 769 ms 595 ms
#
------- End of Forwarded Messagereschly@BRL.MIL ("Robert J. Reschly Jr.") (03/20/89)
Mike, I am not convinced that hanging your hat on rfc1009 provides an out (see page 2 of rfc791, expanded upon on page 14 and repeated on page 30). The key point is (as stated on page 2) "If the time to live reaches zero before the internet datagram reaches its destination, the internet datagram is destroyed." Since this field must be decremented as a part of the packet processing (also stated in rfc791), you need to check whether the outgoing TTL has reached zero. Later, Bob
satz@CISCO.COM (03/20/89)
Mike, If the Butterfly Gateways support LSR or SSR and if I accidently screw up my source routing, will the packet ever die? In other words, I am asking if I can construct a source routed packet within the DDN which could live forever. Greg Satz cisco