A start on the User Manual



> There are couple of problems with HTML--first, no HTML viewer support
> vector graphics (such as EPS) as far as I know, and generally bitmaps
> are just awful. My eyes start to ache if I try to read some HTML
> document where math is in ugly bitmaps (althought I think that
> some versions of latex2html can generate antialiased bitmaps which
> is much better but still awful).

You have a few examples of math formulas in fifth to seventh figures at: 

http://www.univ-orleans.fr/EXT/ASTEX/astex/doc/fr/nouv30a/html/nouv30a.htm

which are quite readable I think ? In my implementation of TeX4ht ( software by 
E. Gurari to transform Plain TeX or LaTeX2.09/2e to html) I use dvidot (by E. 
Mattes) to generate automatically formulas as bitmaps, and it is even better 
(although I have not put any example on the web).

The criterion "beautiful / not awful" is maybe important for artwork, but for 
mathematics, in my opinion, the criterion for displaying math on the web ought to 
be "exact". Formulas available as bitmaps cannot be modified and will always be 
displayed correctly as created originally by the author by any browser, provided 
only it is able to display correctly bitmaps. 

Formulas available as parts of pdf files, with symbols taken from vector fonts 
are displayed certainly more beautifully, but they can be displayed incorrectly 
on the user's machine even if it has been displayed correctly on the machine of 
the author, for a multitude of reasons. With pdf for example, a paper of mine was 
printed correctly with Acrobat Reader 3, but with AR 4 some letters were not 
printed. Or a character can be substituted to another one, for example A can be 
displayed instead of  greek alpha, il the font used on the user's machine is 
bugged or ill-installed.

This might not be too dramatic for pure text, and for not too important text 
(say, personal letters, internal notes in a Univ dept etc.), as the mind can 
correct errors in general (as one can understand mistyped words from the 
context). But in maths, it would be catastrophic if a formula written (a+b)^2 by 
the author and displayed as such on his machine, is displayed (a+b) on the 
machine of user X, and (alpha+b)^2 on machine of user Y, because there is no way 
for the mind to correct a formula, except by refinding it, which may be very 
long.

In short : with pdf and vector fonts, there is no way to guarantee that a 
formula, displayed correctly for somebody, will be displayed correctly for 
somebody else at some other time. Especially for Acrobat Reader, it is not free 
software and there is no way for the scientific community to control its quality. 
The past experience seems to show that each version brings its own set of bugs, 
in an unpredicatble way and with no guarantee that they will be corrected.

If one wants to stick with pdf files, it seems to me that the only reasonable way 
to have mathematics articles that can be read safely in the long term (for 
articles posted in public archives such at arxiv.org), is using bitmap fonts, as 
generated by Dvips+Ghostscript. Acrobat Reader is very poor at displaying such 
fonts, display is ugly, and one has to use Ghoscript to display these pdf files 
in a readable way.

> Secondly, HTML tends to be
> (althought it isn't necessary) splitted in hundreds of small files,
> from which some thing is extremely hard to find.

Well, html does not tend to anything, it is the authors that do tend to splitting 
documents in hundreds of small files, to allow fast loading and browsing. But as 
you say, there is no necessity : you can generate from a single LaTeX or texinfo 
file both types: large number of small files, adapted to fast load and browsing, 
and one unique huge file adapted to text search. In the GNU project, both types 
are available, for (almost?) all docs, so you have the choice. With PDF, usually 
you have only the opportunity to load a big file. So I would say html is more 
versatile.

> Not to mention
> the fact that very rare browsers support antialiased text while
> almost all PDF readers do.

The necessity for antialiasing with pdf files comes from its limitation:  you 
cannot increase the size of characters displayed, because the width of text is 
fixed, and to read a line with big characters, you have to zoom and then use 
scrollbar to read each line, if the line is not entirely displayed on your screen 
- whic is very uncomfortable.

There is no necessity with html, as lines are wrapped automatically in the 
window. html is issued from sgml, so you ought not to have the size of a font 
embedded in a html file : if some text is displayed in an illegible manner on 
your machine, the html solution is to display it with a larger font, adapted to 
the size of your screen and the state of your eyes. There is no need of 
antialiasing, whose aim is to improve display of text at a given size.


Michel.Lavaud@univ-orleans.fr
http://www.univ-orleans.fr/EXT/ASTEX
ftp://ftp.univ-orleans.fr/pub/tex/PC/AsTeX
liste de discussion: astex@univ-orleans.fr
Abonnement à la liste: envoyer un message de contenu
"sub astex Nom Prenom Etablissement" à sympa@univ-orleans.fr