(Sorry, I will not be able to participate in the 6man discussion at the
meeting, as I have to be in another session.)
It appears likely that it is impossible to meet all of the relevant
constraints. But less us not pretend that solutions likely this are
obviously sufficient.
One of the major drivers of the UDP checksum question is tunnel
protocols where the tunneling is being handled by high end routers with
large volumes of traffic.
1) Many of such routers are designed such that it is difficult,
expensive, or sometimes even impossible to calculate a per-byte checksum
across the entire encapsualted packet when encapsuatling a packet for
tunneling.
2) When there is high volume, it is clearly desirable that the resulting
packets be able to be load balanced using the existing Internet core
hardware to spread packets across equal cost paths of link groups, while
keeping individual packet streams to single paths. Currently, this is
support for UDP and TCP packets, but not for SCTP or UDP-Lite.
(Effective traffic distribution across link groups is very important in
real deployments.)
2') There is an argument that the right way to handle this balancing for
IPv6 would be to use the flow label. But that is not the current
practice in either the use of the label field or the implementation of
the balancing logic.
I understand that there is a reasonable concern that if such tunnel
packets are redirected due to packet corruption they may get sent to
devices which do not understand what is happening, and may be
potentially confused if we allow the non-use of UDP checksums.
If that is indeed the primary problem, then solutions like negotiation
are addressing secondary issues. (Of course, not everyone may agree as
to what the "primary" problem is.)
Yours,
Joel M. Halpern
Gorry Fairhurst wrote:
Using UDP-Lite would be fine. I'm not sure how a transition to UDP with
no checksum would help the transport concerns with mis-delivery, etc.
Gorry
Margaret Wasserman wrote:
I had a thought on the use of zero UDP checksums in IPv6...
What if we allowed the use of zero checksums for UDP as a _negotiated
option_ in IETF tunneling protocols? Those protocols could default to
the use of UDP Lite (or UDP with non-zero checksums if their designers
prefer) and negotiate the use of Zero UDP checksums when both sides
support it.
Margaret
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
