Why not? Leave the choice of identities to the publisher, then *we*
don't
have to pick any one (or two). We can pick the whole set.
people who carefully read andy's email on "Identities and Authorization"
will see this paragraph:
1) We have been given a list of 5 or 6 identities from which to pick.
If we did not narrow the list, where would most of the complexity be
located? A sender needs to implement only one, but would a receiving
MTA need to implement an authorization path for all, and might there be
overlap or conflict?
put more plainly: allowing multiple identities (or policy algorithms)
places an exponential burden on the receiver as they must support for
all possible combinations.
No. Why? I can't see that the "receiver" has to support *any*. The
publisher isn't making a law here. The receiver is free to choose which (if
any) scope they want to evaluate in. If I as a receiver chose to implement
a scheme which evaluates authorisation based on the MAIL FROM identity, but
the publisher asserts that a record is to be applied to some other
identity, I can return an undef or even apply it against the publishers
recommendation (I'd be foolish in the 2nd case). If a publisher asserts
that a record covers all possible "identities", I can chose to apply it to
only one (or none).
if the base number is small, the number of
choices remains small and the burden isn't that onerous. if the base
number isn't small, then the number of choices ensures that the solution
will never see widespread deployment.
That's futurology (right next to astrology).
Which "solution" requires that all published identities must be considered
together in all their combinations? Where is the requirement that any
receiver adopt such a "solution"? There is no "burden" on the receiver save
that which the receiver choses.
A broad base for the producer does not require that all consumers will be
interested in all possible assertions. Some might, some will not be.