Hi again. There has been a flurry of messages on that (obtuse ? not obtuse ?)
subject, and now silence again. But I am afraid nothing is resolved...
First, it is not possible to let the january text stay as it was. Following
strictly what's in it, nasty things can and will happen (there was nothing in
my 'bad journey' example that did not follow exactly the letter of that text).
After all the things I have read on the subject, I still maintain that the
changes I proposed in my message of feb. 13 are a possible correct solution.
1- These changes do not document something I invented, but rather formalize
what was explained to me by more competent people. The best explanation,
the one that made me understand the matter, was from Mark Crispin.
2- The problem is *not* with quoted-printable only. It is more a question
of how texts must be encoded, be it in quoted-printable or in base64.
The point is that texts must be in CRLF representation before the
encoding (the encoding itself being in the representation expected by
the local MTA - whatever this one is). One could argue (and some have)
that this is trivial, since RFC822 says that CRLF is the normal form
for texts. But it really is not as, for example, the message from
Nathaniel dated feb. 24, shows clearly.
3- While my 'proposal' includes modifications to the definition of quoted-
printable, the purpose of those modifications is really to remove from
it any traces of portable line ends. This is needed to make base64 and
quoted-printable idempotent, as stated everywhere.
So the changes do two things :
- state clearly that texts must be in CRLF form before any encoding, something
that is very very easy to overlook, even if it can be deduced from RFC822.
- remove portable line ends from quoted-printable, and correct the dispositions
in it that were contrary to the first point.
We can not let incorrect things find their way in RFC-MIME, and then count
on a further 'requirements' document to fix them... /AF