Erik Naggum on SGML and DSSSL

From:         Erik Naggum 
Date:         1996/07/29
Subject:      Re: New horizons for Scheme
Message-Id:   <3047629064455260@arcana.naggum.no>
Newsgroups:   comp.lang.scheme

  ----------------------------------------------------------

[Patrick Stickler]

|   Having tracked the DSSSL standard from its inception, and having worked
|   in the SGML field for going on 8 years, I'm here to tell you that DSSSL
|   will certainly take the commercial publishing world over (eventually),
|   and probably the Web as well.  As for the "annoyance" of dealing with
|   SGML, I'm willing to wager that either (a) you've not done much work
|   with SGML or (b) you were applying it to a problem for which it was
|   either ill suited or overkill.  I personally find SGML to make my life
|   *much* easier and to remove orders of magnitude more annoyances than it
|   adds.

you seem to need creds: I have been working with SGML since 1989, I
maintained the SGML Repository at the University of Oslo for nearly 6
years, I was a member of the ISO working group for nearly 5 years, and I
have contributed some 1500 articles to comp.text.sgml over the same years.

my main goal in working with SGML was to find ways to represent information
for arbitrary reuse (_one_ of which is printing) -- some means to let us
humans encode information such that programs could deal with some of the
complexity of the intuitive structures in our communication.  my goal was
most emphatically _not_ to put ink on paper or pixels on screens, and that
is the only place where SGML has been widely used.  over these years, my
annoyances with SGML grew, and when I had gotten three chapters into a book
on SGML (working title: "A Conceptual Introduction to SGML"), I realized
that I had been willing to accept such a staggeringly large number of
annoyances and flaws in this language that I could neither write further on
my book nor tell people in any other way that SGML is the Right Thing,
because it isn't.

SGML does have some good ideas, but they are neither exploited nor explored
by the user community, and they drown in useless garbage that make the
reuse of information coded in SGML harder than any other language that
could be used.  SGML's only actual forte is that it is a standard.  the
standard itself is horrendously complex and amazingly lacking in precision,
making it an ordeal to study it and understand it _completely_, which about
a dozen people did last time I checked (at the beginning of this year).

after having invested nearly 5 years of my life in SGML, I decided that it
had no future outside of print shops and government contracts, and I could
either waste my time and money not getting any of the necessary changes
through the ISO "review" process, help people encode a structure in their
information where they had even less clue and incentive to understand their
own information than they do with databases, or do something fun and
exciting and rewarding.  I chose the latter.

I have an unfinished article in http://www.naggum.no/~erik/sgml/.  it is
not particularly pedagogical, nor has it been polished at all.  three
posters I presented at SGML '94 are under ftp://ftp.naggum.no/pub/SGML94.

|   I agree, though, that DSSSL is a good example of an ISO standard "done
|   right".

DSSSL is perhaps the single best standard I have seen come out of
ISO/JTC 1/SC 18 or SC 22, and I have seen a lot of them.  it is unfortunate
that it was wasted on SGML.  however, one may hope that DSSSL spawns some
serious redesign in SGML over the next decade.  until then, read DSSSL to
find out about a very elegant solution to very complicated problems, and
design your own "input language".  I have my own ideas, of course, and will
write them down during this fall.

#\Erik