Regarding publishing TXT records, Ted Hardie (speaking just for himself)
I am concerned that MUST here means that we can never transition.
While I would agree with shifting SPF2 from MUST to SHOULD,
I believe, speaking personally, that strengthening MAY to MUST
goes to far. Both as SHOULDs makes some sense; at the moment
someone can deploy without SPF2 if they cannot publish
the new record, but in the future someone can deploy without
publishing a TXT record (which they may want to do in order to
use TXT for something else).
A few points:
1. If we went with "MUST" like I proposed, a future transition away
from TXT records takes the following direction:
a. Initially, everyone publishes TXT records, and those that can
publish SPF2 records. Those that can, query SPF2 records, and
then TXT records if no SPF2 record exists. (Those who can't
query SPF2 records just query TXT records.)
b. Time passes, and more and more people gain the ability to
publish and query SPF2 records.
c. Eventually, almost everyone publishes SPF2 records and almost
everyone queries for SPF2 and gets a response. At this time,
the TXT records still exist but are not being queried for.
d. A future standards effort reduces the MUSTs about TXT records
to MAY (or possibly even to SHOULD NOT).
2. Even if the spec said SHOULD about TXT records instead of MUST
(as Ted suggests), the future evolution is exactly the same.
3. If time passes and SPF2 records do not become widely deployed /
then MUST was probably the right thing for the spec.
4. While I still believe that "MUST" is a better choice, I could
probably live with "SHOULD".
-- Jim Lyon