In <m0nEoR3-001IDCC(_at_)garnet(_dot_)msen(_dot_)com>, Ed wrote:
The other useful thing to do would be to poll implementors and see how
far they have gotten to shipping the Microsoft RTF format around, and
determine whether there's existing practice that has been shown to
work that's suitable for writing up in an RFC.
I have a copy of Microsoft's RTF specification in front of me.
I'd be interested to hear other people's opinions, but mine is
that it is only marginally adequate as a personal computer
document interchange format. I would be loathe to see it
introduced as any kind of Internet standard.
(Disclaimer: though I have read the RTF spec carefully, I have
not tried to implement any of it, so the following opinions may
be unduly pessimistic.)
The problem with RTF is, unfortunately, quite common: the people
designing it only thought they were coming up with something
which was independent of any particular word processor. It
embodies numerous assumptions which are particular to Microsoft
Word, and which would be significantly difficult to map to any
other application. I would actually be surprised if many other
word processor vendors were supporting this format. (I'm not
very familiar with the market, though, so my feeling could easily
be wrong.)
It would be very appropriate to develop and make widely available
a program to convert RTF into the eventual MIME rich text format.
I would hate to see all MIME readers having to parse and implement
RTF, though.
Conceptually, RTF is similar to SGML in that there is a fairly
general syntax which delimits various types of entities. The
problem is that the semantics of too many of the entity
descriptions are far too idiosyncratic to Microsoft Word.
(One of the recent complaints about RFC1341 richtext could also
be applied to RTF: several parts of the spec are quite imprecise;
an implementor would need access to an RTF application,
preferably one of Microsoft's, with which to experiment to
resolve fine points.)
Another problem with RTF is that it encodes characters using the
character set of the originating machine, which theoretically
requires receivers to know about several machine-specific
character sets (IBM-PC, Macintosh, etc.). A MIME RTF
content-type could mandate a particular character set, but that
would preclude direct word-processor-to-mail operation by users
with the wrong kind of personal computer.
Microsoft distributes paper copies of the RTF spec free for the
asking (they even pay postage). The address to write to is
Microsoft Corporation
RTF/Applications
16011 NE 36th Way
Box 97017
Redmond, WA 98073-9717
I could write up a more detailed review of RTF if anyone is
interested.
Steve Summit
scs(_at_)adam(_dot_)mit(_dot_)edu