For what it's worth, I made the following design decisions in my
richtext-to-groff converter:
* bold, italic, underline, and fixed pitch are separate attributes that can
be turned on or off independently. There are only two levels of any of
these -- off or on -- so "<bold><bold>foo</bold></bold>" is the same as
"<bold>foo</bold>". Also, anytime you leave an environment, the old
environment is restored, so "<bold><bold>foo</bold>bar</bold>" is the same
as <bold>foobar</bold>.
* <bigger> and <smaller> are inverses, that is,
"<bigger><smaller>foo</smaller></bigger>" looks the same as "foo".
* <subscript> and <superscript> affect character size as well as the
baseline. (However, Dana's argument for making them separate is worth
thinking about.)
Other issues not raised recently:
* the only line breaks that "count" are those indicated by <nl>.
Other line breaks are treated the same as spaces. This means that
<center>this
is
one
line</center>
appears as a single line, while
<center>this<nl>is<nl>four<nl>lines</center>
appears as four lines.
* changing left or right margin (via <flushleft>, <flushright>, <center>,
<indent>, or <indentright>) causes a break in running text.
My understanding is that richtext isn't really designed to give you a lot
more capability than the average crt terminal, and so issues like multiple
levels of bold and underline aren't apropos to what it was designed for.
It's probably a mistake to try to extend richtext to be that flexible.
However, I do think the richtext spec needs some tightening up for the next
go-around.
Keith