Looking back to the first 10 years of the Web, it certainly looks as if there has been a revolution. The Web now runs on HTML, HTTP and URLs, none of which existed before the '90s. But it isn't just because of the quality of these new formats and protocols that the Web took off. In fact, the original HTTP was a worse protocol than, e.g., Gopher or FTP in its capabilities, and HTML back then also wasn't quite what it is now: no embedded images, no tables, no colors...
So in the early days many people had their home pages on FTP servers rather than HTTP servers. And that fact shows nicely what made the Web possible at all: it didn't try to replace things that already worked, it only added new modules, that fit in the existing infrastructure. HTML can be served by FTP or Gopher servers; browsers can display plain text or FTP directory listings; URLs allow for all kinds of protocols, etc.
And nowadays (the year 2000), it may look like everything is XML and HTTP, but that impression is only because the "old" stuff is so well integrated that you forget about it: there is no replacement for e-mail or Usenet, for JPEG or MPEG, and many other essential parts of the Web.
Of course, technologies may get replaced by better solutions over time. GIF is slowly being replaced by PNG, SMIL replaces several other formats, ditto for SVG. The Web is in constant evolution. There is no "day 0" at which everything starts from scratch. At any time there are old and young technologies working together.
It has to be like that. Technologies improve at different speeds. Trying to release even three or four new specifications in sync is already stretching our capacities. And throwing away software that works, although imperfectly, and teaching everybody something new would be a huge waste of resources.
There is, unfortunately, a tendency in every standards organization, W3C not excluded, to replace everything that was created by others with things developed in-house. It is the not-invented-here syndrome, a feeling that things that were not developed "for the Web" are somehow inferior. And that "we" can do better than "them." But even if that is true, maybe the improvement still isn't worth spending a working group's resources on.
And especially, don't design a technology to work only with things that haven't proven themselves yet, no matter how promising they sound. The risk is too high. For example, at the moment (2000), XLink promises to become a powerful technology for adding hyperlinks to new XML formats, but CSS3, which is currently being developed and which has to style hyperlinks in XML documents cannot rely on XLink. It has to deal with non-Xlink hyperlinks as well.