Le 28 juil. 09 à 13:19, Francis Dupont a écrit :
In your previous mail you wrote:
To take in consideration your (useful) remark, the proposal
could be
improved by replacing a SHOULD by a MAY:
- IPv6 hosts MAY accept UDP zero checksums (but of course MUST
still
send non-zero UDP checksums).
- v4 to v6 translators MAY either compute UDP checksums when
receiving UDP zero checksums OR translate with the zero checksum
unchanged.
=> if you believe this non-issue still needs a solution in standards
IMHO it is far better to deprecate zero UDP checksums.
Different view here.
- Deprecating a permitted behavior doesn't change existing hosts in
any short time.
- Relaxing an unnecessary constraint lets host free to take advantage
of it or not.
If you don't
believe just reuse RFC 2765 choice.
RFC 2765 says
Fragmented IPv4 UDP packets that do not contain a UDP checksum (i.e.
the UDP checksum field is zero) ... will not be translated by the
translator.
The (improved again) proposal would be that two additional behaviors
are permitted (with a MAY).
They are:
- translate with the zero checksum unchanged
- Compute the UDP checksum (in the single fragment case where it
consumes CPU power but is not complex).
Regards,
RD
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------