>>have struggled with how at the same time to keep things readable and
>>also remain consistent with the spirit (if not always the letter) of
My feeling is that rational viewers of structured text should interpret,
not display, the markup. There is no need to surround text with <em>
</>. Instead, the system can select a typographic emphasis, e.g.
italic. There is no need to show <ed> </> on text which says ``ed and
supplemented by Robert A. Kraft''. etc., etc.
A competent SGML viewer will allow searching and computation on the
tags even though not normally displayed, or allow them to be
selectively displayed in the event the reader is not satisfied with
the interpretation which the presentation represents. Thus, in the
face of visual ambiguity you can still find what you want. For
example, suppose the viewer adopts the position that all foreign words
should be italicized. This makes it impossible to distinguish one
language from another, or from, say, emphasis. But the viewer should
make it easy for you to search for the next French word, bypassing all
the emphasized text (if it contains no French...).
SGML is supposed to take no position on presentation but only on
structure. The idea that humans, instead of software, should be
reading the markup is probably a perversion of SGML's historical
intent and in any case a terrible idea.
The plea I heard two years ago from linguists was that they were
constrained to low cost, low performance computers. Now low cost no
longer implies low performance, and consumers of SGML---or any
markup---should require human readability of their viewing technology.
This is well within the capability of document processing systems like
those of Frame and Interleaf running on hardware which this year costs
$1800 in the U.S.
Dept. of Math. and C.S.