I don't have anything to add to Mark's and Nathaniel's remarks on the
"declare SMTP to have improved functionality" issue. That issue has
always been understood as having the implication of identifying SMTP
servers that don't have the "improved functionality" as somehow
defective and substandard, and it is from that reasoning that the
"declare...broken" terminology originated (although I strongly believed
that Robert had used the latter language).
It may be that David is suggesting something different at this point;
if so, the message is still not clear to me.
The best solution from an IETF standpoint is probably to ignore these
suggestions and get back to work on RFC-ZZZZ or its offspring. That
work, if standardized, will clearly identify the sending of anything
other than 7bit characters, in octets with leading zero bits, in a line-
and character-oriented transmission model, as inappropriate for SMTP on
the TCP/IP Internet in the absence of explicit negotiation between
sender and receiver/server.
Individual implementers may, of course, interpret the combination of
the robustness doctrine and octets with the the high bit set as they
find useful.
But it seems that David has proposed something else this time, which, in
principle, stands apart from the "broken" debate. If we go back nine or
ten months, there seemed to be an emerging consensus on the list that
many problems could be solved if we could settle on a single normative
character set, presumably DIS 10646. Now I always had misgivings about
that plan quite aside from concern that we not standardize something for
the Internet based on a draft that might change significantly. But that
doesn't make it a bad plan, at least in principle.
Now that "make 10646 normative" approach collapsed for several
reasons (at least in my recollection):
-- the DIS crashed and burned in ISO voting, reinforcing insecurities
about using the thing prematurely.
-- the proposed approach was to use compaction method 5 level 2.
That compaction method is strongly ISO8859-1-centric and, hence, subject
to accusations of being Eurocentric with less-desirable representations
for non-European characters.
-- compaction method 5 level 2 was also one of the victims of the ISO
DIS vote, so the proposal's foundation suddenly disappeared out from
under it.
-- First-DIS 10646 omitted a number of important characters and
languages that were felt to be important and that certainly undermined
its claim to universality.
-- A series of issues arose about fixed-width characters, different
uses of control spaces, and the "one glyph, one position" model, most of
them under the name "UNICODE". Those issues considerably muddied the
waters.
We did, however, salvage something very important from the excursion
into 10646-land, and that is the first-level registry structure for
mnemonic.
David has proposed, I think, that we look at the 10646 AUC proposal as a
foundation for a different coding model in the the 7bit world (whatever
it might mean in "other" worlds). Putting my reluctance to get involved
with the unstandardized and partially undocumented aside for the moment,
it would have a nice relationship with mnemonic (since it is basically
10646) and might have a useful relationship to (or be a useful
alternative for character data to) quoted-printable.
Insofar as people believe that, a few years down the road, lots of data
communications will use some standardized revision of the 10646 drafts
and/or that it will provide the common language for conversions among
character sets, this suggestion may deserve a careful and serious look,
quite independent of the question of how that information is transported
over a (real or putative) 8bit connection.
--john