[ietf-dkim] Alternative text for semantics of multiple signatures

[Arvel Hathcock]

 > The MUST NOT was there in the earlier proposal because the association
 > between p= and the headers was by hash values. This proposal removes
 > that, and MUST NOT is not needed. If we use "SHOULD NOT", we need to
 > say when it is OK to do it anyway. Proposal: "To avoid deleting
 > information that might be valuable to the recipient, signers SHOULD
 > NOT remove..."

Yes.  With that change and the previous "SHOULD" becoming "should" as 
discussed by Stephen and Mike I think this text is very good.

-- 
Arvel