I've said it before, and I'll say it again: richtext will only happen
on a large scale when richtext editors happen or translators from
common wordprocessor formats. As a sender of mail I am not going to
type <bold> to attract your attention, I'll use *emphasis*. (And
please let's not reiterate the suggestion that an even simpler format
of richtext could borrow such conventions.)
Therefore the most important restriction for richtext is not that it
is easy to write but that it is easy to create parsers (and to a
lesser extend generators) for it. This is accomplished (although a
better spec of what certain richtext constructs *mean* precisely would
have made it even easier).
It *may* also be important that it is somewhat human-readable in its
raw form. Another restriction is that it should be translatable to
ASCII without loss of too much information, since ASCII is what
(during a *long* transition period) the majority of mail readers will
have to work with (e.g. I'm writing this at home on a VT100 clone and
while I may have an X workstation at work this is not going to happen
soon at home...).
So I think we should be very careful with adding notations for "funny"
characters (no offense meant), and in fact I second Nathaniel's motion
for leaving character set switching out in the second try (I thought
that had been taken out of the draft long ago???).
I have one other remark regarding the choice of ISE 10646 encoding.
Currently my saved mail files share a common encoding with all other
text files on my system. I would like to keep it that way. I think
that this means that I'd like mail to use the same encoding that my
vendor will pick once they adopt 10646. So we should argue about what
character set is best for vendors to adopt (I'd hope all vendors will
use the same one -- will it be Unicode?) rather than about what
character set is best for mail exchange. (My vote goes to UTF-2 since
it is the best way to keep the big investment in ASCII safe.)
--Guido van Rossum, CWI, Amsterdam
<Guido(_dot_)van(_dot_)Rossum(_at_)cwi(_dot_)nl>
(I ran out of funny .sigs)