There is the issue of self-canibalization: since Silverlight is now Microsofts pendant to Flash and these media engines support vector graphics, video, animation, etc., it becomes a question of whether to support both fronts equally (Silverlight and the
browser) or let one stay ahead of the other. In a sense, if all browsers started supporting vector graphics and video, that might dampen the evolution of these plugged-in engines.
It's probably fair to say that Microsoft prefers the Silverlight architechture to the W3C jigsaw puzzle, otherwise I'd expect them to fully support SVG, SMIL, etc. There's another reason for this preference and that's the natural glide path from Silverlight
into WPF with Visual Studio and Expression Blend being the premier development tools for these APIs.
On the other hand it's probably also fair to say that a certain lesson has been learned in that it is not wise to lag too far behind other browsers. So there will be some form of co-evolution, probably - with Silverlight being the clear priority...
In fairness, IE supports or supported native vector graphics via VML and? HTML+TIME - although maybe not in the IE8 engine; I heard CSS expressions (I forget the precise name) were also phased out - but they were proprietary anyway.
I see plugins as a playground for new developments that can be rapidly deployed to the Web (depending on the update experience of the particular plugin).
I believe you state the specific weakness of the w3 specs.
HTML 4.0x xhtml 1.0, plus mathml plus svg plus smil do not work together.
SVG and HTML are top level languages (which in some instances can become sub languages.)
The issue becomes how does the browser interact with all of those specs? on the same page in the multiple domains.
Opera so far has gone the furtherst in implementing a browser that handles it. But the primary browser needs to be XML engine that handles the poorly speced 4.0x. if it starts as a html page. or handle html and mathml islands if it starts as svg.
Since css hasn't been completed let to work in that domain it gets more interesting.
That said I believe that the more correct approach would be to define the underlying browser technologies. that handle any language to implement on it.
IOW xHTML, SVG MATHML would all sub class under a super canvas class. But I doubt we wil ever see such a beast coming from the highly political design by committee world of w3.
IMO HTML 5 is already a failure although it will take us closer to actually defining the browser and its fallbacks.