SGML on the Web

The best thing about this book is that it comes with a copy of the SGML browser SoftQuad Panorama PRO 2.0. The worst thing is that it then takes a great deal of persistence and ingenuity to get Panorama to work with your own SGML documents. Nowhere does the book make obvious how to create the essential associations between SGML documents, their Document Type Definitions (DTDs) and their Style Sheets.

However, once you do manage to suss it all out and get things working, you can begin to experiment with the fascinating world of document structures, documents created from text fragments, automatically generated tables of contents, bi-directional and one to many links, and other features that expose some of the impoverishment and compromise of HTML as we currently know it.



SGML stands for Standard Generalised Markup Language. Strictly speaking, SGML is not a language but a meta-language for defining markup languages. For example HTML is defined in SGML. However, SGML can also be used to create any user-defined markup language such as tags for a <Newsletter> that begins with <Title> and <Issue> details, and contains <Editorial>, <Article>, <Announcement> and <Review> items. The set of tags so defined makes up a "Document Type Definition" (DTD) and each document that conforms to it is an instance of that DTD. The absolutely crucial point is that documents are marked up according to structure rather than appearance. This makes fragments of documents available for re-use. Thus a newsletter editor, with appropriate <Newsletter> markup, could automatically extract or combine all the reviews, or any other sections from multiple editions of the newsletter for use elsewhere.

Appearance is defined separately using Style Sheets, and more than one Style Sheet may be associated with each DTD. The newsletter could therefore be presented in different forms for different purposes, such as for print, for screen, in public and private versions, in a taster version to attract new readers, and so on. In effect a DTD is rather like a database schema, and markup like field names.

Much of the book consists of 40 worked examples which take "small steps" through SGML in gently increasing complexity. Many of the examples relate to the DTD for HTML, although features not present in HTML are also covered. The text and examples also appear on the CD- ROM, so one can simply start up the Panorama browser and begin to dissect them. However, as the authors suggest, there is far more to be gained by typing in the examples yourself, amending them, and experimenting with your own documents and DTDs.

I eventually succeeded in doing this but only after working out the roles of two files called catalog and entityrc in associating together documents, DTDs and Style Sheets. This took a couple of weeks on and off, but having once overcome the problem I found much of interest. The book also contains an SGML reference section and several appendices of dubious congruity. Another criticism is that the first part of the title (as shown on the book) may be misleading: you need an SGML- capable browser to handle SGML across the web. However the book appears under an alternative title in some lists, as indicated above.

That the roles of catalog and entityrc are not made obvious, together with a vague unfinished feel about the book, could well be a result of the untimely death of Yuri Rubinsky in 1996. Yuri Rubinsky was well known in the SGML and World-Wide Web communities, and co- founder of the Canadian firm SoftQuad, known also for its web authoring products.

Despite the problems I consider the book good value for anyone prepared to learn SGML by working through the material. The W3C working draft of the HTML 4.0 Specification [1] indicates that SGML concepts are likely to be of increasing importance to the Web, so this could be time well spent.