Over the past couple of weeks the discussion on the VRML mailing list has been a back and forth, mostly civil (for a change), interaction about how to move VRML forward. Once of the most interesting aspects has been the "return" of several major influential VRML players who don't post all that frequently these days, (presumably because they are actually doing real work). Much of the discussions have centered around something called VRML TNG, VRML - The Next Generation (apologies to Star Trek).
Chris Marrin arguably the most experience VRML browser builder in existance has put forward a proposal called Emma which as Rikk Carey one of the editors of the VRML ISO standard says is pouring gasoline on the discussions.
Marrin's arguments are well thought and cogently stated, and I will oversimplify them as follows: VRML is too complex and some of it's features are too interdependent. The interdependencies make writing a conformant browser too difficult. The event model, the order of evaluation of events is important for conformance yet difficult or impossible to guarantee according to the VRML spec.
All of these arguments may be true. The question is what to do to fix it. The choices presented by various people boil down to "start from scratch" or "support everything". An additional approach, presented by some, is that of depricating some features as a new spec is evolved.
The supposed crisis instigating all of these developments are a perceived lack of tons of VRML applications, and the turmoil of the two main VRML browser companies. Both Intervista (WorldView) and CosmoSoftware (CosmoPlayer) have effectively vanished from the VRML scene, at least temporarily. The reality however is that Intervista, as has been publicincompatibilitiesally announced a couple of months ago, is now part of Platinum Technologies, and the VRML portion (at least) is also part of Platinum. Soon (we presume) this will all become clearer. At least in the not-to-distant-future we won't have to worry about incompatabilities between those two browsers!
It seems to me that the issues of concern are ones of process. How do we fix VRML without throwing out the proverbial baby (dancing I assume) with the bathwater. The VRML Consortium has a remarkably open process established. If you have a good idea put together a working group and do it. As a few have stated starting from scratch is much more fun than fixing the old. Quite frankly I'm not competent enought to know if VRML can be "fixed" but my gut feeling is that it can, and that a process for evolving it can be defined without scrapping everything.
I'd like to finish by repeating a posting by Mark Pesce (he's always so good to quote) about a recent experience. He was helping a computer novice friend get up on the net with a new Win98 machine. The got the ISP connected and started to surf, they saw some Shockwave and it worked, saw some Java it worked, and saw some VRML and it worked! VRML was already there, installed and invisible....real ubiquity.