Gavin Thomas Nicol said:

> The fact is that anyone with a reasonable amount of SGML experience
> ended up using a core subset similar to XML. Ultimately there was
> little new in XML, because it was based on something with a fairly
> long history. Some things, like I18N and explicit DTD-less support
> were good additions.

I think there were three factions in SGML: those who used OmniMark, those
who used SGMLS or NSGMLS, and those who had to roll their own tools. While
people in the first two factions certainly sometimes normalized their data
into fully-unminimized forms, it really was the roll-your-own crowd,
notably browser makers, who needed something simpler than SGML.

SGML is alive and well in some quarters, XML-DEVers may be surprised to
hear. Last month I had to do some S1000D related work
(military/aerospace), and it seems that SGML is pretty entrenched there
still, for the moment.

It is also worth reflecting that in some sense XML has failed, to the
extent that XML was an effort to deny that people needed reduced markup:
reduced markup formats have thrived in the form of Wikis and even the
dreaded SML/YAML/CVS/.ini files. XML's adoption has meant that Wiki
formats are not specified using SGML, and so there is no standard way of
extracting an XML-compatible informations set. XML has made a lot of
information available with a standard infoset, but also alienated a lot of
information that potentially could have been accessed with the same
infoset.

Cheers
Rick Jelliffe