SGML on the Desktop


SGML on the Desktop

Seminar given at the Beyond HTML workshop, University of Glasgow, 21 July 1998, under the auspices of :TLTSN and :HATII.

You can take a copy of the .overheads for the talk.

If you're interested, you can also look at the .SGML source of this document, the .DTD it uses, or even at the .output of NSGMLS (*format), to see what the formatters are given to play with.

The other seminars today have given you some idea of the power of SGML, and of the range of applications it can support. My task now is to zoom in on the finer details of what tools you need to assemble to transform your ambitious (I hope) plans into files on a server, or ink on paper.

First the bad news: I can't tell you `If you want to get a desktop SGML system working by tomorrow, go out and buy X'. I can't tell you that because no such thing exists. SGML has hitherto been confined to very large documentation projects with the resources to develop custom software. The canonical example is the maintenance manual for a jumbo jet, which will have hundreds of different authors, and numerous different views depending on whether the reader is an engineering manager or a mechanic. This has meant that tools have been designed to be very powerful, rather than easy to use.

With the emergence of XML, tools will appear which encourage SGML to percolate down to much smaller projects (though I recently read a passing remark which referred to a `small document' one with fewer than 50 authors), and perhaps within a couple of years we might see off-the-shelf systems appearing.

However, to be brutally frank, to get an SGML system going you really do need either money or expertise: both would be nice, but you're not going to get terribly far with neither. Having said that, an extremely powerful SGML system can be built with entirely free software (or as Rutherford said, talking about the early successes of the Cavendish Lab: `we had no money so we had to be clever').

Because they're rather complicated tools, working in a relatively small market, SGML tools do tend to be rather expensive. I expect this will change as XML tools become more common.

So, I'm not going to say `buy this' or `buy that', but instead explain what the various components of an SGML system are, explain how they relate to each other, and hint at what you might consider when you put your system together. In short, I'm not going to give you any answers here, just tell you the right questions to be asking yourself.


Norman Gray
21 July 1998