I think this statement gives dangerously wide latitude for intermediate
systems to damage end-to-end-ness. It seems to me that a router should
only do something outside fundamental routing behaviour when this has
been explicitly approved, either through protocol negotiation or through
manual configuration, by sufficiently many affected parties that the
others can't tell that anything out of the ordinary is happening.
Where do you think this leaves rewriting DSCPs at network
boundaries?
But anyway, the document is a statement of principle rather
than a standard, and I'm not sure that it's possible to be
absolute about these things when the interests of the user
may be counter to the interests of the network
administrator. That's not quite true - it is possible to be
absolute. But I don't think it's likely to produce good
results or progress on what really are some rather difficult
and sometimes subtle questions about what happens between a
sender and a receiver.
Melinda