ietf-822
[Top] [All Lists]

richtext <nl> model considered harmful

1992-02-03 12:44:51
As most of you are painfully aware by now, I've been playing a lot with
richtext, and I've become convinced that we got one thing wrong.  The
whole idea of richtext, you'll recall, was to have an extremely simple
markup model for enhanced text that was as readable as possible in its
raw form, for the benefit of users of non-MIME-compliant mail readers.

I am increasingly convinced that the way richtext handles newlines does
not meet the latter constraint.  As I've gradually refined the
Andrew-to-richtext algorithm, there are two things that I am still
getting a lot of complaints about:  excerpts and newlines.  The excerpt
problem is very Andrew-specific, and says nothing interesting about
richtext, and anyway I've got an idea for fixing it.  The newline
problem, however, is more fundamental, and therefore of more general
interest.  The problem is that people find it quite annoying to see
"<nl>" scattered liberally throughout messages they read.

Here are two other models for how richtext might handle newlines:  

1.  It could say that all newlines are REAL.   A long line of
text-to-be-justified is then represented, in richtext, as a single long
line, implying that most people will use an encoding, probably
quoted-printable, to send such text through the mail.  This implies that
instead of a lot of "<nl>" markers, people viewing raw richtext will see
a lot of lines that end with spurious "=" characters.  I think that
would be better, but perhaps not enough better.

2.  It could say that any sequence of N newlines represents N-1 real
newlines.  Thus a lone newline is just "white space", but two
consecutive newlines are a real line break, and three are a real pair of
newlines, e.g. a paragraph break.  (This, by the way, is the Andrew
model.)  It is very readable -- most people won't even notice it in most
contexts -- but may be the death knell for our claim that richtext is
SGML compatible.

Personally, I vote for solution 2, although there is probably a
procedural issue with changing ANYTHING at this late date.  Opinions,
anyone?  -- Nathaniel