SO how is inter-operation hurt or hindered by this specification's
making the check normative vs. non-normative?
Some food for thought on this question:
One of the results returned by ADSP is "domain does not exist" (see
section 3.2). If you predicate one of the outcomes of the protocol upon
an optional "it might be a good idea to do a lookup, you know, if you
feel like it" then what you'll have are different implementations of the
protocol each able to truthfully claim compliance yet each returning a
different result _for the same input data_. Bad.
This is what I have been referring to in earlier posts saying that ADSP
results should be "reliable" and is why, if we're stuck on the question
of normative vs non-normative, either normative must win or we must
remove "the domain does not exist" from the list of ADSP results (which
creates all the problems my side in this debate has been preaching about
in other posts).
Hence, the wisdom of Wietse's solution which is to have the spec assert
that ADSP is to be applied only to Author Domains which exist in DNS.
This nicely eliminates the need for language like "you MUST (or SHOULD)
do a DNS check" thus satisfying (one would hope) one side of the debate
while at the same time requiring the acquisition of the data which
"domain does not exist" needs - thus satisfying the other side of the
debate.
It's beautiful. I note with great concern that my opponents in this
debate have not commented on Wietse's proposal.
Arvel
_______________________________________________
NOTE WELL: This list operates according to
http://mipassoc.org/dkim/ietf-list-rules.html