Rich Fox's proposal

[mogul (Jeffrey Mogul)]

I think I will assert my perogative as chair of this working group
(something that I would rather not do too often) to draw the line
here.  I don't think our charter extends to allow us to meddle with the
design of existing transport protocols (i.e., TCP).  This is distinct
from any suggestions we produce for the implementation of transport
protocols.

In other words, it is OK for us to say "this is information shared by
the TCP module and the IP module in a host's implementation," but it is
not OK for us to say "this is how the protocol defined by RFC793 is to
be changed".  The former presents far smaller issues of compatibility
and bureaucracy.

Clearly, we are going to have to change the IP protocol in some way; at
the very least, by defining a new option if not a new header flag.  I
regard this as within our charter, since it would be absurd to charter
a group such as ours without leaving us this option.  (Chris and I had
suggested an ICMP-only mechanism in our paper, but it has not been
warmly received.)

Having drawn this line, I do not want to entirely discourage proposals
of this kind.  I think they can be a good starting point for thinking
about more feasible approaches, but I would hope that they are made as
such, and that the proposers not be too upset when these proposals are
discarded.

-Jeff