Re: "Fragmentation considered harmful"

mogul (Jeffrey Mogul) Wed, 31 January 1990 03:20 UTC

Received: by (5.54.5/4.7.34) id AA20385; Tue, 30 Jan 90 19:20:23 PST
From: mogul (Jeffrey Mogul)
Message-Id: <>
Date: 30 Jan 1990 1920-PST (Tuesday)
To: Philippe Prindeville <philipp@Gipsi.Gipsi.Fr>
Cc: MTU Discovery <mtudwg>
Subject: Re: "Fragmentation considered harmful"
In-Reply-To: Philippe Prindeville <philipp@Gipsi.Gipsi.Fr> / Wed, 31 Jan 90 02:48:36 -0100. <9001310148.AA08050@gipsi.Gipsi.Fr>

    Going over the DECWRL report (not the ACM paper) I was thinking
    that the IPMP protocol is a good idea, and it offers the chance
    to kill several birds with one stone by collecting a lot more
    than just the MTU along a path.  Additionally, I would suggest
    collecting security levels available and type-of-service levels
    available.  Perhaps a space for route recording too, since an IP
    option is limited to 7.
Thanks for the praise!  (For those of you who don't know, IPMP
stands for Internet Probe Message Protocol.  We made it parallel to,
rather than part of, ICMP, so that gateways wouldn't have to parse
every ICMP packet that went by.)  However, I suggest that given
the current direction of the MTU discovery working group, it might
be better to consider the IPMP in a different WG, and more along
the lines of a management aid or an augmentation to the routing
protocols.  Someone could volunteer to JNC to head a new Working

If enough people request it (mail to me, not the list) I will
post the IPMP protocol proposal to this list, but that will be
the end of it.

	Some might argue it is a security breach to let one
	level convey information about (the existance of)
	other levels.  I will leave that to the spooks to

I think you're confusing protocol archictectural layering with
multi-level security.  Any correctly-implemented multilevel
security system wouldn't let this information leak out, via
IPMP or otherwise, if it shouldn't be allowed to leak.