As a note, converting HTML 2.x to HTML 4 + Embedded CSS or XHTML + Embedded CSS is actually quite easy. Just create style names to replace each of the major HTML 2.x tags then use those. Most modern browsers run HTML 2.x through a similar script tied into a quirks CSS/JS loader document.
I think for more reliable rendering someone should write a tool that does this conversion for web page authors.
An HTML 1-2 to XHTML/CSS Core strict converter would be a godsend. This would go double if it could add SVG embedding of images for single page publication and then have a script to break it up into multiple pages for parallel loading once on the server.
Another thing that would be cool is a simplified HTML similar to Coffeescript is to Javascript that is easily converted by the appropriate XHTML/CSS loader page.