Re: SMDS MTU

mogul (Jeffrey Mogul) Fri, 01 June 1990 17:38 UTC

Received: by acetes.pa.dec.com (5.54.5/4.7.34) id AA01057; Fri, 1 Jun 90 10:38:37 PDT
From: mogul (Jeffrey Mogul)
Message-Id: <9006011738.AA01057@acetes.pa.dec.com>
Date: 1 Jun 1990 1038-PDT (Friday)
To: kzm@nms.hls.com (Keith McCloghrie)
Cc: mtudwg
Subject: Re: SMDS MTU
In-Reply-To: kzm@nms.hls.com (Keith McCloghrie) / Thu, 31 May 90 23:15:05 PDT. <9006010615.AA14764@nms.>

    Don't know whether you saw this, but of interest is the
    SMDS MTU of 9188.  Can we assume that by the time this is
    available, that all routers will have implemented the 
    modified (for MTU-discovery) ICMP message which contains 
    the MTU size.  Else, it won't do much for the plateau 
    algorithm.
    
I noticed this too, but I don't think it's a serious problem,
for several reasons:

(1) The current table has a plateau at 8166 octets, which is
something like 12% off from 9188 ... so the damage would be
slight (remember, a 12% difference in packet size will cause
much less than 12% difference in actual throughput).

(2) Right now, there is only one MTU larger than 9188 that
is known to me (the 16 Mbit Token ring's maximum possible MTU).
That means that only in the case where someone connects such
a LAN directly to an SMDS network will there be any issue at
all.  In all other cases, the SMDS MTU won't matter, because
the sender's initial estimate will always be lower.  [Does
anyone know if anyone actually runs token rings with an MTU
of 17914 octets?]

(3) It's probably true that by the time anyone builds an IP router
that supports SMDS, router vendors will know enough about MTU
Discovery to judge whether they should implement it (i.e., it
might not be a standard by then, but its destiny should be clear).

-Jeff