"Tony" == Tony Finch <dot(_at_)dotat(_dot_)at> writes:
Tony> No, it's one that doesn't contain a valid email address.
I really don't see how you can interpret the phrase 'non-empty' as
meaning 'containing a valid e-mail address'.
Malformed header fields are addressed by step 5; if the field body did
not contain a valid email address, this would constitute a malformed
header field, and the algorithm would not return any PRA.
It's important that two implementations of the PRA algorithm never
return two different answers as to the PRA, even in the presence of a
malformed messages. (It is acceptable that one implementation returns
an answer and another does not).
What you propose is dangerous, because it depends on the
interpretation of a 'valid e-mail address'; specifically, liberal
implementations typically choose to accept addresses that are not
strictly valid according to RRFC2822.
If two implementations have a different opinion as to the validity of
an e-mail address, that could result in them picking different
headers. (This is precicely the flaw in the originial PRA algorithm
in the Caller ID draft)
The effect of this is that even if the users MUA is Sender ID aware,
and displays the PRA, there's no guarantee it is displaying the same
identity that was validated by the MTA.
-roy