I'm also not sure how much your complaint applies to HTML 5 since that was originated by browser vendors wanting to standardize the new stuff they were already implementing and to reverse engineer and then compatibly implement how old stuff was implemented in Internet Explorer.
Anyone still remember VRML? The attempt to bring 3D graphics to web browsers in 1997? Netscape browsers supported VRML and you could actually view 3D models in it, written in XML.
I did my MSc project using VRML and Java, At the time most of the implementations were missing large parts of the VRML2.0 spec. It had some nice stuff, like being able to embed a program to create geometry.
Part of the problem was just being a little early with most machines lacking 3D graphics chips, really limiting the type of scene you could create. Today even some peoples phones have 3D chips in them, so that less of a problem.
Having real 3D building without a plugin like live maps currently requires would be nice, I know the mapping provides have the data already. Plus loads of other ideas I've not even thought of. I still feel it will have an uphill battle but really hope I'm wrong