Date: Thu, 30 Jun 2005 18:50:01 -0400
From: Jeffrey Hutzelman <jhutz(_at_)cmu(_dot_)edu>
| Have you read the spec in question?
I have not, and I expressly will not, because that cannot possibly be
| The issue is not that the presence of an unknown option interferes with
| congestion control. The problem is that when the endpoints _do_ implement
| the option, they can sometimes decide to disregard TCP congestion control
| procedures such as slow-start, which while implemented on an end-to-end
| basis nonetheless affect congestion control on all intervening segments.
You mean the network dies if end nodes don't do the TCP congestion
control that TCP normally expects to be done? That's wonderful
news for the "bad boy" brigade I would have thought. How many days
from now before they have TCP implementations that ignore congestion
control, and by so doing, apparently, totally cripple the net?
They don't need a new option to ignore congestion control, they
just adjust their TCP stack code a little. In fact, they'd do
better (from their point of view) without using a new option - that
would make the packets readily identifiable, and easy to filter, if
they were doing harm that anyone could detect.
Or could it perhaps just be that the net might survive this, it seems
to survive UDP packet streams from time to time, and as best I recall
it, there isn't a lot of congestion control (or feedback even) in UDP.
Ietf mailing list