Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site alice.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!allegra!alice!trickey
From: trickey@alice.UucP (Howard Trickey)
Newsgroups: net.text
Subject: Re: embedded-command text systems
Message-ID: <4681@alice.UUCP>
Date: Sun, 8-Dec-85 11:45:23 EST
Article-I.D.: alice.4681
Posted: Sun Dec  8 11:45:23 1985
Date-Received: Mon, 9-Dec-85 03:24:34 EST
References: <471@harvard.ARPA> <773@mmintl.UUCP> <734@tpvax.fluke.UUCP>
Organization: Bell Labs, Murray Hill
Lines: 61

Please don't judge TeX by the Ullman Database  book.  One problem is that
it was set with TeX78, the predecessor to the current TeX (so, as I
pointed out in net.internat, the hyphenation algorithm isn't the Liang
one).  And it uses older Computer Modern fonts.  Knuth has just
finished a complete touch-up of Computer Modern, and well-known font
designers have participated, resulting in "much improvement", in their
opinion.  But the biggest problem is that full-frontal TeX leaves too
many graphic design choices to the writer.  Ullman just mimiced the
style he was used to with troff -ms.

Knuth recognizes the need for competent book designers, and used one
to design "The TeXbook".  That is probably the best example of the use of
TeX available in bookstores, though it too used the old fonts.
For the average user, the expense and bother of hiring a designer means
that it won't be done.  There, a system like LaTeX should be used.  That
is a Scribe-like interface to TeX designed by Leslie Lamport, and as in
Scribe, the user is supposed to describe only the logical structure of
the document, leaving it up to expert "design-style writers" to make
the graphic-design decisions.  Lamport had some help from book design
experts in making up his default styles, though the world would be
enriched if more design styles vetted by book designers made it into
the LaTeX library.  The LaTeX manual published by Addison-Wesley is
an example of its use.  I don't think it is totally successful, since
the examples are laid out two-columm (input|output) with a resulting
distracting changes of margin size.

As for the debate about WYSIWYG:  I agree with those who say that
it is possible to build a WYSIWYG system with the logical description
capabilities of LaTeX or Scribe, but we aren't there yet and until
we are, I'll stick with LaTeX.  Interleaf is almost there, but the
current lack of (1) automatic numbering of things and the ability
to symbolically cross-reference; and (2) decent math setting, is fatal.
Also, the nicities like inter-character kerning and by-paragraph
line breaking that Tex provides were not there the last time I looked.
I haven't looked at whether it is possible to add one's own logical
structuring categories to Interleaf, but it better damn well be possible.

Seeing what you're getting is mainly important to the extent that
you are doing the graphic design yourself, since you want immediate
feedback as to whether you made the right choice.  With a system
where the user isn't supposed to make such decisions (and, as I
just argued, naive users shouldn't), the ability to see stuff
immediately is a sort of instant gratification that falls mainly
in the category of frill.  Unfortunately, this isn't quite true,
since the user still might want to reword things to avoid bad
line and page breaks.  Knuth would spend several such rewording
passes when doing chapters for his books, and he would have
appreciated a system that would try to keep up with previewing
his document in one window as he edited in another.  But he
didn't feel the need for seeing how the font changes or math
would turn out --- one quickly learns from experience what to
expect.

Maybe one of these days I'll do such a two-window TeX system.
I think the need for massive computing power can be minimized
by keeping a sort of intermediate form of the document around
from previous runs, and checksums of the "environment" at various
points.

			Howard Trickey, At&T Bell Labs
			research!trickey