Kurt Cagle has posted a summary of his keynote speech from SVG Open 2005. He covers a lot of ground on XML and the next generation of GUI. Pretty inspiring stuff as an SVG enthusiast. The amazing part is that even though we've all been hearing about subscribing to software since the 90's, this time I can see it coming. Not soon like next month, but soon like in a couple years. It's not going to be exactly AJAX and it's not an ActiveX control. It'll creep up on us, in big leaps and little improvements. We'll be subscribing to software before we realize we're doing it. Just like a lot of expectations that were overhyped the first time around, it's just taking a lot longer than originally expected.
Note also that the model for making money off this stuff won't be what it was originally, either. People might pay for it directly, or they might not. Right now Firefox extensions and the handful of XUL applications out there are already using this sort of model. They're platform independent, delivered over the web, and can be updated as needed with very little effort on the part of the user. Everyone has their own reason for writing a Firefox extension and the same will continue to happen as subscription software spreads. Individuals and companies will figure out how to make money off of them in time if they haven't already.
I don't expect everything to work out precisely as Kurt's laid it out, but I doubt he does either. That's inherent in any crystal ball gazing. He offers a very realistic view of what can be done with existing technology and what will become much more feasible as implementations of SVG, E4X, XForms and other enabling standards shape up. The co-existence of XAML and SVG is guaranteed - just as XWindows and MS Windows have co-existed forever. There will be translation software, there will be parallel (perhaps often redundant) development. There will be open source and closed source. The browser has become a viable framework for application development. Maybe the next round of PC speed enhancements will just keep us at current speeds but chunking through all that text markup that XML holdouts love to gripe about.