Greg,
I think this is very useful, but, for me at least, it tends to
reinforce the view I was trying to feel out in a message that probably
crossed yours in the redistribution pipe this morning.
I guess the summary of that message, when your taxonomy is applied to
it, is "let's simplify this story by having the sender specify the
conversion directly, rather than working up an intermediary typology
through which the conversion can be deduced.
In that system, your five cases collapse into three: none,
quoted-printable, and base64, with a clear extension to "don't convert"
if anyone but me (and Stef, probably) still think that should be
offered.
I'm also, personally, back to ignoring the "binary, no line length
transport" question. Several reasons have been identified why this is
not a good idea, at least without additonal safeguards. One subset of
the class of safeguards probably requires PEM, which means (as I
understand it) at trip through Base64 anyway, so there is no point
considering those cases as no-line-length binary. It has also been
pointed out that a "binary" message in RFC-XXXX format (or, more
generally, with any form of vaguely RFC822-compliant text headers)
is going to need processing and extraction--it isn't going to be like
FTP or BITNET NETDATA, which can be used "right off the wire" as
received.
Several people have said, more or less "we wanna have binary transport"
and others have said "we know how to do binary transport". I think the
latter is clear, and people are certainly entitled to their preferences.
But, given the objections that have been raised and the fact that this
does make things slightly more complex and increases the number, and
possibly severity, of failure cases (i.e., it is an interoperability
threat), I'd like to see this dropped until some of the advocates can
come forward and say "we *need* binary transport because...", or "the
savings in bandwidth are really important because...".
It is possible, for example, that we could meet the real need for
"binary 8 bit transport" with the "enclave-only, no conversion" model
that the WG decided was inappropriate for "mail". I'd like an
opportunity to think about those cases. That is not an opportunity I
have as long as the "binary requirement" is expressed without technical
justification and a real statement of purpose.
--john