I'd like to add my support for SVG. First, I believe that it is widely used in mobile technology though this is through conversations about 2 years ago rather than seeing code. It was one for the first languages developed by W3C to be used over-the-web and has been consistently supported for the last 10 years. I initially thought it would be a killer-app in 1998 and said so, especially since Adobe then had enthusiastic support.
Because SVG is a full XML dialect it can be easily mixed with other XML applications such as XHTML and MathML. It's possible to include XML information in SVG elements so that documents can be fully semantic (i.e. carry a data-meaning as well as a graphical one).
SVG is effectively feature-complete for a 2-D graphics language. There is no technical reason to invent anything else. It does high-quality rendering, animation and can support interaction through its own elements or through Javascript. It is therefore unlikely that major web software manufacturers will NEED to invent anything else.
The main problem is not that SVG is "dying" - it is not - but that it takes a long time for the various companies to converge. When they do I would expect SVG to emerge as a consensus.
Part of the problem is that graphics are not yet seen as a semantic problem - it's often "easier" to inflict a dumb animation on a client than to set up the infrastructure to send semantics over the web and repurpose at the client. But as the semantic web develops then standards such as SVG will be increasingly important. If, for example, you wish to mash-up geographical information then either you use a properietary solution such as GoogleMaps, or you look for a shared standard approach. You can never count on free proprietary solutions in the long run.
Note also that in many areas which receive public funding there is an increasing drive towards Open standards and this is another reason why SVG is well positioned.
