In <F10C6AA6-FDF0-11D8-B6CF-000A95B3BA44(_at_)hxr(_dot_)us> Andrew Newton
<andy(_at_)hxr(_dot_)us> writes:
Despite lack of agreement on the outcome, there is agreement on the
stated positions. In the words of Olafur:
1) Specify everyone to be compliant at all times.
2) Specify clearly full complaint state and tolerate non-compliant
state during phase-in.
I know the chairs asked to not discuss the IPR issue any more, but the
SenderID license seems to rule out option 2. According to section
1.2, a "Licensed Implementation" must be complaint with all required
portions of the SenderID.
While the SenderID license only directly references marid-pra and
marid-core, the marid-core I-D has a normative reference to the
marid-protocol I-D and section 3 of marid-core requires you to use the
check_host function.
So, if marid-protocol says that check_host MUST query the SPF RR and
an implementation doesn't, then it is not confromant and the
implementation does not qualify to use the SenderID license. If you
are in a situation such as Jim Lyon described at the iterim meeting
where you are in a pure Microsoft network, the Windows API doesn't
even have the ability to query for new RR types, you are kind of out
of luck. As Jim described it, you can't even code up raw UDP packets
because they won't make it out of the network.
The problems of the license requiring conformance was brought up by
Dan Quinlan back in June, if I recall correctly. This is just one
example of such a problem.
So, if we are planning on supporting SenderID, it appears that we MUST
allow everyone to always be conformant at all times. This means that
we MUST NOT require implementations to query SPF RRs.
-wayne