From: Werner Koch [mailto:wk(_at_)gnupg(_dot_)org]
On Tue, 7 Apr 2015 19:38, Neil_Hunsperger(_at_)symantec(_dot_)com said:
One solution is PGP Partitioned format, which supports in-line
signatures for ASCII bodies and detached signatures for other email
IIRC, the Partitioned format has no means to guarantee the integrity of
the entire message and thus you can replace attachments.
To clarify, one can remove attachments, re-arrange them, or change any MIME
property except for the file name. Replacing attachments would require a
signature of the new file made by the same key at about the same time.
Up until the recent spate of new UIs implementing OpenPGP I would have expected
the discussion of encrypted email formats to die out, with the comparatively
simple PGP/MIME format (or perhaps a variant with secure subject lines)
becoming ubiquitous. Now implementers seem to be asking for something that
simultaneously meets the needs of existing back-end message decomposition and
existing front-end usability. The OpenPGP v4 format seems to provide enough
flexibility to solve these use cases so I'd consider tackling OpenPGP v5 and
email as separate tasks with the advantage of having separate timelines.
Using the above as a starting point, what aspect of a MIME signature's
impact is left to solve?
The micalg parameter should be removed or made optional. The micalg
needs to be emitted before the signed data and the signature. It is
useless for OpenPGP and a major hassle for any one-pass creation of
signatures because only the signing tool can determine the used hash
algorithm from the set of signing keys. This should not be
controversial because may MUAs use a fixed string anyway.
+1. PGP Desktop actually used a fixed string until ~5 years ago when Enigmail
reported it was causing signature verification failures.
-Neil
_______________________________________________
openpgp mailing list
openpgp(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/openpgp