By requiring SPF2 records to be looked up when they are not widely
deployed (or at all?) and making TXT record lookups optional when they
are already widely deployed, the specification effectively doubles the
number of DNS queries required for Sender-ID lookups and makes it more
difficult, slower, etc. for SPF queries to succeed, fail, timeout,
etc. in practice.
suggested change 1: Waiting for a negative answer for SPF2 queries
before looking up TXT is not going to yield acceptable performance for
most implementations, this should not be suggested unless the suggestion
is coupled with a flag day.
suggested change 2: Different implementations may have different results
as long as TXT records are deployed. This should be noted.
suggested change 3: Since there is currently a forward path defined for
transitioning from TXT to SPF2, a flag day should be suggested if this
poorly-reasoned pedantic path is taken. Using TXT forever would be much
more practical and cleaner. New additional "TXT" records can be defined
for similar future working-around the difficulty in deploying new DNS RR
types.
suggested change 4: change "later" to "latter" in 3.4.1
--
Daniel Quinlan
http://www.pathname.com/~quinlan/