Well, obviously GPU hardware has become a lot cheaper, more powerful, and more widely available in the last 20 years. VR display hardware hasn't become ubiquitous in the same way, but it's still way beyond what was previously available to consumers.
Also, VRML took the approach of trying to build all kinds of special-purpose functionality into the platform (see e.g. http://gun.teipir.gr/VRML-amgem/spec/part1/concepts.html#Eve... or http://gun.teipir.gr/VRML-amgem/spec/part1/concepts.html#Int...). This is similar to what old-fashioned fixed-function OpenGL used to look like. Over the years, graphics APIs have evolved to become more general-purpose and programmable. And we've seen a similar trend with browser APIs: consider how Service Workers have replaced the old AppCache, or how MediaSource has replaced proprietary video streaming plugins. The design of WebGL/WebVR is a lot more flexible, and better suited to the modern web, than VRML was.
Obviously this doesn't mean that VR is going to be a huge success this time around, but it's also not fair to assume that nothing has changed in the last 20 years.
Also, VRML took the approach of trying to build all kinds of special-purpose functionality into the platform (see e.g. http://gun.teipir.gr/VRML-amgem/spec/part1/concepts.html#Eve... or http://gun.teipir.gr/VRML-amgem/spec/part1/concepts.html#Int...). This is similar to what old-fashioned fixed-function OpenGL used to look like. Over the years, graphics APIs have evolved to become more general-purpose and programmable. And we've seen a similar trend with browser APIs: consider how Service Workers have replaced the old AppCache, or how MediaSource has replaced proprietary video streaming plugins. The design of WebGL/WebVR is a lot more flexible, and better suited to the modern web, than VRML was.
Obviously this doesn't mean that VR is going to be a huge success this time around, but it's also not fair to assume that nothing has changed in the last 20 years.