Heavens - an Orlowski article that I mostly agree with. However, for the sake of consistency I should raise a few token disagreements and comments:
- Yes, SGML was all-singing and all-dancing, and full of rich semantics. I have a copy of the SGML Handbook on my office shelf, and have consulted it within the last three months. Parsing SGML, however, was a royal pain; TimBL's decision to design a simple angle-brackety markup language (I hesitate to call it an SGML application, because it didn't get a DTD until HTML 2.0) was a pragmatic one, and the simplicity (or perhaps paucity) of early HTML has in retrospect been a positive thing; easy to author, and fairly easy to write a bare-bones parser for.
- Your reference to XHL is interesting - you have a better memory than I do. However, XHL owed much to HyTime (which, as an aside, made SGML look almost straightforward by comparison), and much of the XHL work ended up being recycled in XLink. Of course, both XLink 1.0 and 1.1 have raised barely a ripple. Blame the browser manufacturers for being unwilling to support them.
- It's interesting that you don't also criticise the design of HTTP and URIs. Like HTML, their initial versions made some naive assumptions that later versions would correct.
If you're going to criticise the early Web for what it didn't do, you'd do well to consider why it succeeded when other contemporary hypertext systems (Hyper-G, for example) didn't.