At 2:58 PM +0100 11/17/00, Dirk Ooms wrote:
>I have the impression that a lot of the complexity in these chained
>extension headers is caused by the way Source Routing is specified.
The chained headers aren't complex. (Warning: digression ahead...)
The issues under discussion arise from the extra-architectural use of
layer-violating devices along a delivery path. The Flow Label is in
IPv6 specifically to avoid the need for (and the parsing expense/
complexity of ) looking at transport-layer port fields to do flow
classification. Perhaps we ought to have moved the port fields into
the IP header instead, as in Xerox's XNS / Novell's IPX; that was
actually considered and discussed at great length in the early days
of IPv6's predecessor, SIP, and rejected as being too radical a change
to the IP architecture which would have required significant redesign
of IP implementations. And though that would have made it easier to
do port-based firewall filtering for IPv6, it wouldn't have helped all
the other forwarding-path layer-violations that some people deem
desirable today, e.g., stateful firewalls, "performance-enhancing
proxies", "layer 7 switching", etc., but all of those are at odds with
end-to-end encryption, which we were/are hoping will someday become
common in the Internet. Better to put the effort into IP-architecture
-compatible solutions to those other problems.
I'm also surprised that people are using the rationale that speeds are
increasing to argue for changing IPv6 to make it easier for the routers
to do more functions than just forwarding packets. I would think that
ever increasing speeds are a reason to *reduce* the number of tasks that
routers do. (For example, that's why we eliminated a header checksum
from IPv6.)
>If Source Routing would be a hop-by-hop extension and the IPv6
>destination was always the final destination the header mess wouldn't be
>that big (the disadvantage being that Loose Source Routed packets need
>additional procesing in non-fixed hops, but who is using Source Routing
>anyhow?).
Mobile IPv6 is using the Routing header. There's some chance that that
that header will show up in a significant number of IPv6 packets, some day.
(I happen to think it was a mistake to use the Routing header for Mobile
IPv6 -- encapsulation would have been much better -- but it's probably
too late to do anything about that now.)
Steve
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page: http://playground.sun.com/ipng
FTP archive: ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------
