What killed XHTML was the lack of backwards compatibility and especially the first-mover problem with embedding and the ads ecosystem. Namely, making a large pr

What killed XHTML was the lack of backwards compatibility and especially the first-mover problem with embedding and the ads ecosystem. Namely, making a large production website fully validate was very hard, yielded zero ROI, and if anything went even slightly wrong such that invalid XML was emitted that would create a site outage that you couldn't easily detect server side. That's a bitter pill to swallow even as is, but now throw in the common practice of mixing in markup from third parties for ads, social buttons, login forms and other things and now you can have a major site outage caused at random by third parties, which, again, cannot be detected server side. As for RDF, GPT-3 is sort of the ultimate repudiation to that whole line of AI research. GPT-3 doesn't care about structured data at all, it can't even use it any better than a human would.