On Aug 31, 2007, at 2:22 AM, Jari Arkko wrote:
- If a "consenting nodes only" model is adopted, is any arbitrary
behavior by those consenting nodes deemed acceptable by the
non-consenting nodes? Previous discussions seemed to indicate
that at
least some folks would answer "no" to this question.
In other words, how is using RH 253 between consenting nodes (allowed
to behave arbitrarily) any different than allowing consenting
nodes to
run RH0, and non-consenting nodes disabling RH0? One difference
(cynical hat on) is that opponents of source routing now force
proponents to get new functionality implemented and deployed -
only to
have the same debate a few years from now when it comes time to
standardize the functionality.
I think any proposal for special re-forwarding to new destination,
based on RH0, RHx, or anything else would have to be evaluated
by its merits. Some of the questions that would be asked in
review include, for instance, how is consent obtained from
the participants obtained, and how can the consenting nodes
ensure that they are not used as part of an attack (e.g.,
amplification).
Jari
Thanks, Jari.
I was also speaking mostly about the functional aspects of the
mechanism, not the policy aspect. I suppose when the IETF publishes
an RFC (or decides not to deprecate one), it essentially 1. endorses
the described functionality and 2. implies that potentially negative
consequences that could arise if the functionality were widely
deployed have been considered. Using RH 253 among consenting nodes
allows the IETF as a body to remain silent on the merits and
potentially negative consequences of that particular mechanism /
use. That is a difference.
Of course, in this case the deprecation of RH0 essentially
establishes a position on both RH0 and sufficiently similar
functionality. It is sort of like case law, with the RH0 deprecation
draft as the legal opinion. I think this is partly why folks were so
interested that the text in the RH0 deprecation draft focus on the
single issue of most concern (e.g. oscillation DoS), rather than the
merits of source routing in general.
Is this consistent with your thinking?
Thanks,
Dow
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------