The simple notation for even complex mathematics and the diversity of the symbols and characters sets available makes LaTeX the typesetting system of choice in many of the scientific fields. Tens of thousands of articles, theses, and reports have been written in LaTeX and most publishing houses that deal with scientific papers use LaTeX for handling, storing and archiving their documents. Therefore it is to be expected that all these parties wish to protect their investment and prefer not to have to recode their mathematics formulae for hypertext purposes only.
The LaTeX2HTML translator solves the problem of presenting mathematics in HTML by converting each mathematical sentence into a bitmap image. Although simple and straighforward, this approach seems a little unreasonable in general, since in many cases an article of a few pages can generate many hundreds of bitmap images, which have to be stored with the document, kept up to date, and transmitted with the document over the Internet, thus wasting an enormous amount of bandwidth. Therefore, a clear need for a translator from LaTeX mathematics into 's primitive mathematics was considered an important goal. Thanks to the increased displaying capabilities of complyable browsers, most inline mathematics and a fair proportion of display equations can be translated into source code and hence transmitted in textual format together with the rest of the document, doing away with well over 90% of the images that are created in the case where only bitmap images are generated. In addition, mathematics text can be searched for keywords as the rest of the document, thus increasing the value of the HTML document.
The math2html program has been interfaced to the LaTeX2HTML program via a new
option -html3. When this option is specified, LaTeX2HTML will first
pass the LaTeX input source code through the math2html translator. In
this case, native code will be generated for mathematics and
tables when math2html can handle the input. In case math2html cannot parse
the given LaTeX input, it gives an error message and LaTeX2HTML
creates an image as usual.
At CERN we have translated thousands of pages of manuals and hundreds of physics articles. We found that math2html successfully translates on average 95% of all mathematics present in the input files, thus reducing by a substantial amount the number of generated bitmap images.