On Wed, 4 Mar 1992 11:30:31 -0500 (EST) you said:
OK, you're right. I overstated the case a little bit. My real concern
is that PERMITTING 0d0a to represent CRLF is opening a can of worms.
The rest is a red herring.
How about a compromise in which we basically adopt Alain's proposal,
except we keep the old provision that CRLFs must be CRLFs, i.e. they
can't be represented by 0d0a in quoted-printable? I believe that this
will solve all my practical objections. -- Nathaniel
This is of course doable. I am working at a remote site this morning, where
I have net access, but no papers... I'll try to post a new proposal this
afternoon (UTC+1 - should still be morning for most of you).
But one should remember that when q-p will be used to encode binary objects,
any occurence of '0d0a'x will have to be encoded as a line break for
transmission. I may be a little pessimistic, but I fear that a smaller
number of MUA will be correct enough to put back '0d0a'x in the decoded
binary object that if I had encoded =0d=0a. Of course a MUA failing to
do this correctly would be non-conformant. Stay tuned. /AF