On Mon, 2004-11-01 at 12:27, Frank Ellermann wrote:
But syntactically your idea is the same as
Sender-ID for _different_ policies: minimally two records.
Correct.
For _different_ policies your solution uses up to six records:
Yes.
And I grant that positional modifiers can put those into one record.
If your scope includes mailfrom, add an "op" modifier to
your record.
That's a new idea, please use another name for it to avoid any
confusion. The "op=" was never meant to be required, not in
the "other protocols" version (with an op=any), not in "other
tests" (ot=), and not in "other properties" (without op=any).
I meant to propose that:
v=spf1/mfrom,pra stuff
be considered invalid, so the above would be rewritten as:
v=spf1 op=pra stuff
Which is compatible with "op=" as discussed elsewhere. (Which is why I
used the same name. If I'm wrong, I'll switch, no problem.)
The upshot of all this is that if we have a non-positional "op="
modifier, we could then also have scoping with "v=spf1/pra,pra2 " syntax
without the need for an spf1 version bump.
Avoiding the need for that version bump lessons the marketing
distraction/confusion of spf1 vs. spf2.0, eliminating any perceived need
of publishers to "opt out" of pra scope with an spf2.0 placeholder
record.
(I'm basically assuming the existence of an option to say "v=spf1/pra "
for pra-only records would mean that MS could stop pushing the confusing
spf2.0 format, *and* it would effectively get us Unified at the same
time, because they could push either "v=spf1 op=pra blah" for the normal
case, or a "v=spf1 blah1", "v=spf1/pra blah2" for the case where the
records should be separate. No need to make a confusing version change
just to push their agenda, and it would be less likely that such a
document could go on the standards track too, were there to exist an
option to do the same thing that didn't require a version change.)
--
Mark Shewmaker
mark(_at_)primefactor(_dot_)com