More and more content is being taken entirely off the open web and siloed behind a server that talks an unstable proprietary protocol, with exactly one blob of javascript in existence that knows how to tunnel requests over HTTP to access shreds of that content and cram them into an utterly non-semantic DOM. We are hurtling backwards into the client-server hell the web had saved us from.
Yeah, I don't see that. I see more and more accessible APIs[1] and pages having more and more an incentive to being semantic due to search engines now reading that data (hRecipe, for example).
Service architecture have also been moving from stuff like SOAP to REST, which is definitively more open and accessible.
And even Ajax-ladden webpages are still just a Firebug Network tab away since they all run over HTTP, and then you have a nicely structured data format instead of having to deal with messy HTML pages.
A JSON (or SOAP) backend is only usable by third parties if its API is kept stable. There are far too many devs who redesign their backend request and response formats at the drop of a hat because they think their js client is the only one that matters (a self-fulfilling prophesy) and they can replace it simultaneously. And their responses tend to look like "here's some more markup to stuff into an arbitrary location in the DOM we're using today", not semantically structured (e.g., Rails now has this built into JavaScriptGenerator). A given site can be reverse-engineered, but anything built on that is going to be fragile and short-lived, much more so than when the typical visual rendering desired for a page determined its structure.
I don't see how that's worse than unstable, non semantically structured HTML markup of yore. In the worst cases, we're not really worse, and we have much more semantic content nowadays.
