Like the LATEX form of the index, the HTML version is also constructed completely automatically. The same index data as before is fed into makeindex, but this time, a different makeindex style file is used, in order to produce HTML output.
Because many World-Wide Web users have slow Internet connections, and many of the journal indexes are large, I decided to split each HTML index into multiple files. Also, rather than provide separate author and title indexes, which would require the user to search in multiple places, I decided to offer only a single collective index for each bibliography file, covering all of the words in BibTEX abstract , annote , author , classification , keywords , note , subject , and title values.
The top-level index file contains pointers to each of the sectional files, plus additional data and commentary that is extracted from the BibTEX file comment header. There is usually one sectional file for each letter of the alphabet, another for mathematical text, and possibly yet another for special symbols.
Each sectional file contains index entries where the composite page numbers are all hypertext links. Because the sectional files are generated sequentially, it is not possible to provide pointers to the other sections, like those in the top-level file. Instead, there is just a single hypertext link to the top-level directory index.
Each index entry hypertext link points to a single small file (average size about 5KB) containing just a single bibliographic entry in BibTEX form. To aid in navigation in the index, the top-level directory index is repeated in this file.
Following the BibTEX entry, there is a mini-index labelled Related entries ; it contains index entries for all of the words in the BibTEX entry that are found elsewhere in the index. This makes it very easy to locate other entries in the bibliography that may be related to the current one. Consequently, if a title word is absent from the mini-index, then the reader can be certain that it does not occur anywhere else in the bibliography.
All of the HTML files produced for the index conform strictly to the current HTML grammar level, and have been prettyprinted to ensure a uniform format.