The PRA algorithm (deliberately) specifies its behaviour for messages
that are not conformant with RFC822 or RFC2822.
I think it would be desirable to clarify why this is (ie to ensure a
deterministic result in the presence of liberal parsers) and what the
implications are (specifically that we are not in any way requiring
implementations to accept messages that they would otherwise have
rejected as malformed).
I suggest adding the following note to the end of section 2 of -pra.
Although the algorithm specifies how messages that are not in
conformance with the provisions of RFC822 or RFC2822 should be
treated for the purposes of determining the purported responsible
address, this should not be taken as requiring or recommending that
any systems accept such messages when they would not otherwise have
done so.
However, if a liberal implementation that chooses to accept messages
not in conformance with RFC822 or RFC2822, then it MUST use this
algorithm to determine the purported responsible address of such
messages.
-roy