Evidently you can handle all of MVC within flex, but I am still seeing it primarily as the presentation layer — perhaps as my familiarity increases, I will be more comfortable letting it handle more. in the meantime, however, it is easier for me to see flex as a way to present data — and only parts of it.
I plan to use SWFObject to include flex components within a larger page — and that also lets me add the content directly to the page, to be swapped out by the richer component if it is supported. That seems to be a decent solution for search as well as acessibility. Perhaps mobile as well?
Now, I have never been a fan of alternate presentation based on user agents, etc (browser sniffing, etc). Build to standards for a more robust, consistent solution. Otherwise there are more maintenance and consistency concerns (not to mention the fact that you can offend users by offering them up ‘dumbed down’ content). That being said, I do like the idea of progressive enhancement- which is how I am choosing to see this. I don’t want to deprive users of important information, but I do want to offer the best user experience to those who can support it.
What I’d like to figure out is how to ensure the same content is being served up, regardless of the presentational format. So if we’re feeding XML into the flex component, I want the same content written to the page. I did some quick prototype work using XSLT at LN, I’d like to investigate building modules not simply of flex, but of the entire package: XML (data), being rendered on the HTML page and supplemented as possible in a richer format.
I think something like a Flash renderkit for JSF would be both great and in-line with your post above. The content would all be in/served through the xhtml (Facelets) and the end result would be based on the renderkit, which could be driven by user-agent support.