Subject: Re: perhaps time to check our TCP against spec?
To: Kevin M. Lahey <email@example.com>
From: Dennis Ferguson <firstname.lastname@example.org>
Date: 04/07/1998 17:20:48
> My enthusiasm for the new in_maxmtu MSS scheme (aside from
> the fact that it comes straight from RFC 1191) comes from
> my perception of the reality of multihomed hosts, asymmetric links,
> and mobility. I think that it is very likely that a connection
> could be issued from the Ethernet interface of Host A,
> and the return path could come back over the FDDI interface.
> If we advertise only the Ethernet MTU, we'll never get to
> use the FDDI MTU, even if all traffic goes that way.
> And while this doesn't seem very likely in the LAN,
> it sure seems possible on the Internet (based on Vern Paxson's
> recent stuff on routing asymmetry).
I think this is a case where one's expectations of `normal' differ
depending on whether the box is a host or a router. If the box is
a router it is absolutely the case that the best MSS to advertise is
that derived from the interface with the largest MTU. Not only are
asymmetric paths possible in this case, but it is also possible that
route changes over the duration of the connection may move the incoming
(and outgoing) path from the ethernet to the fddi and back.
Hosts, on the other hand, generally do not have a way to move traffic
between interfaces like this. That is, if the source address included
in the outgoing packets is that of the ethernet interface then all return
packets will almost certainly return via that interface; hosts don't
generally support the mechanism (i.e. a routing protocol) that would
be needed for neighbours on the FDDI ring to recognize that the host's
FDDI interface is a path back to its ethernet address (one could contrive
examples by hand configuration to force the asymmetry to occur, but this
is not `normal').
Thus a host which sources packets with the address of an ethernet interface
can be almost guaranteed that return packets will arrive via that interface.
And while there is nothing strictly wrong with advertising any MSS you
care to, it does violate the `be conservative in what you send' principle
if you advertise a FDDI-sized MSS in a situation where you can be pretty
confident that all incoming responses will arrive via an ethernet, PMTU
discovery not withstanding.
There is an issue of how the box would know if it was a host or router,
but this is not a unique problem. Hosts and routers do different things
with ICMP redirects as well, for example.