David Shaw wrote:
On Wed, Mar 17, 2004 at 10:54:14AM -0500, Ian Grigg wrote:
[settling the end-of-line issue for 0x01 signatures]
Myself, I'm minded to say it should be either:
i. there should be no trimming, as the binary
(as opposed to cleartext) signature form
implies careful handling (mime/zip) and
thus special care is less needed,
OR
ii. it should be the same as the cleartext
signature method, for less confusion.
(Unfortunately, both these require a compatibility
change.)
I'm not sure that the change will hurt compatibility. Given the way
that PGP and GnuPG handle 0x01 signatures (by including their own
variation of canonicalized text as part of the message), we could pick
either of these with no significant problems. Note that PGP currently
does (i), and GnuPG currently does (ii), and they manage to
interoperate most of the time. (Not to say that there aren't other
implementations out there).
I would be perfectly content with either solution, with perhaps a
slight leaning towards (i) due to a desire to not mess with user
supplied input unless we have to.
Right now, I'm inclined towards (i) myself as
a signing implementation could warn whether
there are trailing spaces, and a verifying
implemetation could try to trim spaces as
a fallback technique.
iang