I was reading the April issue of Release 1.0, Esther Dyson's punditry
journal, and came across a description of SGML as a small, extensible
language, a tool for creating rich data languages, and I was reminded
(for no discernable reason) of Interviews' Idraw tool.
Interviews is a C++ toolkit for X Windows, and one of its associated
applications is a structured graphics editor, called Idraw. Idraw plays
a nice trick with its document format. It writes out its data in the
standard way you see with most structured drawing tools: there's a few
lines which define some global document features (background color,
size), then a line or two for each type of graphic object, giving things
like bbox, color, corners, line thicknesses, etc. The kicker is that
this document format is a subset of PostScript, and they also throw a
header in the file that defines macros so that the document defines a
valid PostScript document -- it can be sent to a printer, and what gets
printed is the document you drew. But Idraw just skips over the
PostScript prologue, and just looks at the descriptions of the graphic
elements -- a *tiny* tiny subset of PostScript. What a win! The
document format used by the tool happens to be useful for things which
have no notion of Idraw, but which speak PostScript!
Of course, this relates to Richmail as well. Suppose that Richmail was
a tiny, tiny subset of some "outer" markup language, say LaTeX
(restricted to just the forms that Nathaniel specified for Richmail),
and was sent with a LaTeX prologue, identified with a content-type such
as "Content-Type: latex/richmail", or some such. Simple
formatters/strippers could just skip the LaTeX prologue (OK, this
probably adds a couple of lines to Nathaniel's 38 -- so now we're at
40), and deal with the restricted LaTeX as a "Richmail" format. People
who don't have Richmail, or want to store this document outside of the
mail world, would have a valid LaTeX document, for which there are
"public-domain" interpreters.
What a win! Document punning!
Of course, it's immaterial whether SGML or LaTeX or even Andrew is
chosen as the "outer" format, just so long as it's some kind of standard
with existing tools for it, with a syntax that maps nicely to
Nathaniel's Richmail syntax.
Bill