SGML: The Bad Stuff

[Previous] [Next]

SGML hasn't been able keep up with the Web. It was designed in an era of slow, expensive computers (by current standards). To make the most of primitive systems, the developers of SGML designed a complex set of minimization features—features designed to make files as small as possible. This minimization makes SGML complex to process and expensive to implement, and makes parsers large and slow.

To create an SGML document, a user must have in place a DTD that describes the structure of the information. This means that the user must go through an expensive information discovery process and define the information's structure in no uncertain terms before creating the first document. The user employs the DTD again and again whenever the document is processed. You will see later that it is overkill to verify that a document's structure is correct unless it or the underlying structural description (DTD) has changed.

Another problem is that SGML is owned by academics, who seem to be more interested in purity than usability. During two quinquennial (five-year) reviews in 1991 and 1996, this group had a chance to simplify SGML for the Web. Many in the industry called for "SGML Lite"—a drastically reduced syntax that would fit nicely in a Web browser. The ISO committee refused to make changes that would slim down SGML.

We needed something that was portable, cheap, fast, and easy like HTML, yet extensible like SGML. And we needed something that was compatible with what we already knew so that we could use existing techniques and tools.