The HyperReader script attempts to read Bookreader format documents and present them as hypertext documents. Sometimes it is very successful, other times not! The reasons are few but significant:

• Bookreader documents are not plain text.
  They are not even marked-up text (in the same sense as HTML documents). Essentially they are a series of images pasted onto an X window. Most of the images happen to be from fonts, but even then characters are not handled as text. There is essentially no information within the file on the documentary function of any text in a book. This may be one reason for the poor performance of the internal Bookreader print functionality (it must adopt a similar mapping then "images" into PostScript "pages").

• Bookreader format is not public.
  At least not that I could find. Hence considerable effort at reverse- engineering the structure (most originally during January 1992 fortunately, I wouldn't have the patience, time or interest now!) In any case, the structure is complex and I don't pretend to understand it all ... so the server tends to break occasionally.

• Development time available.
  Zero. However not being able to get to the wealth of Bookreader documentation from the hypertext enviroment in some fashion (no matter how primitive) seemed too great a deficiency not to make some effort. Corollary: the server is not as fully developed (or debugged) as might be desired.

Limitations

• Client/Server Use.
  Bookreader documents were never really designed to be accessed using client/server technology. There are some "book opening" overheads best done only once (as with the original Bookreader application). With the HTTP (hypertext) environment these must be performed with each access, sometimes introducing noticable latency.

• Mapping of document text.
  The server attempts to map a bit-addressed format (Bookreader) onto a character-cell format (text-page). Bookreader formatting permits parts of a page to be generated non-sequentially, for example to place normal text, then place bolded text within it. This can occur on a per-line, per- section or per-page basis. So the entire "page" for a hypertext enviroment must be created by mapping into a "page" in memory. Sometimes this mapping is not perfect and character strings get garbled or poorly proportioned.

• Fixed-Font presentation.
  This is due to two requirements:
  1. The mapping from bit-addressed to cell-addressed described above requires a fixed character-cell matrix.
  2. Maintaining correct layout for examples, tables, etc. As the Bookreader format specifies the positioning of the document's components, rather than information on their documentary function, relative location must be maintained, generally eliminating the use of the more aesthetically pleasing proportional fonts.

• Image conversion and compression.
  Figures and graphics within Bookreader documents are stored as single-plane bitmaps. HyperReader supports these using the GIF format. Figure positioning on a page can sometimes be clumsy due to the character-cell layout.

• Page layout.
  Apart from the text-mapping process mentioned above, page layout can sometimes be a little odd, particularly if graphics are involved. Some massaging of the resulting page occurs. This often involves educated guesswork. For instance, single blank lines preceding a line beginning with a lower-case alphabetic are most often an artifact of the mapping process rather than intended line breaks, and can be eliminated to produce a better layout. Similarly, bulleted and numbered lists can be enhanced by ensuring a blank line exists between elements. Unfortunately there are not many instances where the original intent can be clearly identified.

Finally ...

Even considering all the limitations of the approach it generally works remarkably well, and does also give access to Bookreader documentation from a character-cell terminal (via a browser such as Lynx). I hope its useful.

HyperReader v0.8;  April 1998