Vincent's latest message was, as usual, thoughtful and reasonable. It
also worries me.
The question that remains, I think, is a simple one: how many
troublesome cases like Vincent's example will there really be, and how
important are they? Without much evidence to support it, I have a gut
feeling that the answer is "not too many and not too important."
My gut feeling is that, if this is carried to its logical extreme,
the answer would be "perhaps very important, and a great many".
Consider the multipliers involved: if you accept "nroff" as motivating a
content type, recall that nroff is the product of one operating system
and a relatively small number of things that have cloned its (nroff's)
code. But fairly low-level formatting languages exist at roughly a
one-one ratio to operating systems, plus or minus a few, so...
and so one and so forth.
This makes a very good place to consider how far one wants to go, and
I'd refer people back to Stef's recent message for what I consider a
persuasive case. There is a critter called ODA. It is designed for the
handling and representation of structured, rich text format, documents
that contain combinations of variant character sets and/or, e.g.,
special format codings. IETF management, in its infinite wisdom, has
created a WG on ODA over Internet mail.
I would suggest that this is a reasonable place to draw the line and
that, if the ODA types are not rich enough to handle nroff and 8859-1,
we send people off to the ISO WGs and make that happen.
In other words, for both this case, and for other cases where carefully-
structured text is needed (maybe imbedded in-line funny glyphs, per
Risto's example, maybe even ordered content sections), we introduce one
additional content type, ODA, and then let those folks do their thing.
The question shouldn't be whether we can diddle Content-type to handle
every imaginable case (we are collectively certainly smart enough to do
that), but how to figure out where to stop. I think Vincent's example
should be used to open debate on the question. Stef thinks Risto's
example (and my comments on it) should be used to open debate on the
question. We are probably both right, not necessarily about the
conclusion, but that a little meta-level consideration of whether it is
desirable and necessary to make this sort of thing work as part of
Content-type is in order before working on the engineering details of
how to do it.
Curiously, being able to say "that is an ODA problem", when appropriate,
meets Nathaniel's criterion about not opening cans of worms: it
involves taking the closed can, worms and all, and passing it to someone
else with instructions that it is their problem. If ODA is adequate
today (it may not be), "they" will have very docile worms when they do
open the can.