Yahoo! Architect and upcoming AJAXWorld 2008 East keynoter Douglas Crockford has caused a stir by writing about how, while the web has grown up from a document delivery system to an application delivery system,  the browser has not kept pace, so that as Crockford puts it "there are now new proprietary platforms from Adobe and Microsoft and others that are hoping to replace the web."

Against this background, Crockford argues, XML itself is jeopardized. "Is XML on the web trending up or trending down?" Crockford asks, before answering:

"Clearly, it is trending down. For data transfer applications, XML is losing ground to JSON because JSON is simply a better data transfer format. And XHTML has failed to displace HTML in the marketplace. The benefit of clientside validation has proven to not be a benefit."

"Douglas,

I noticed the other day that Ruby had crested about August 2006 in terms of the number of citations it was receiving in the press, and has been declining at a rate of roughly 2-3% per month ever since. Given that Ruby is perhaps one of the largest single producer/consumer networks of JSON, it may be worth spending some time looking seriously at whether in fact the arguments you are making are not in fact as applicable to that environment.

Most syndication that I see on the web is XML based, though since its usually called RSS2 or Atom people tend to discount how pervasive that is; the entire SOA stack is XML based, and I'd estimate that something like 65%-80% of all web development currently involves XML at some point in the production pipeline, if not necessarily the point connecting the server to the browser. While there is a fairly significant amount of JSON being flung around in the web 2.0 space, I find that neither JSON transformations nor JSON schemas have really managed to gain much traction.

Is JSON better than XForms? That's an apples and oranges argument. I am willing to predict that XForms + XQuery will become a powerful enterprise model for rich form content, because the enterprise is considerably more XML-centric than the consumer stack is. I'm willing to predict that JSON feeds to most mashups will likely end up being a mix of XML (primarily via feeds) and JSON for quite some time, though I'm inclined to suspect that AtomPub will likely tilt the balance of power towards XML in the long run.

One final note - I suspect that if you look at rate of growth, both JSON and XML are on sigmoid curves, with XML perhaps about 75% of the way along its trajectory, while JSON's probably about 50% of the way along its. Both will continue to rise in usage for quite some time, but both will also reach a plateau point, rather than one replacing te other. Big variables yet to be answered is to what degree is JSON catching on in the mobile space (by most accounts, mobile developers prefer declarative markup), to what extent will mashups continue to rise (my sense is that they are in fact stabilizing or even beginning to fall as the market becomes saturated there, while syndication feeds become the dominant services architecture), and whether a lightweight XML format such as e4x or LINQ gets adopted by other platforms (which I suspect will likely be the case with an ES4 adoption, which you're also opposed to).

JSON's not going to go away, nor should it - there are actually quite a few niches where it is in fact preferable to working with XML. However, I also think that its a little early to predicting the demise of XML on the web ... especially since I see some of the most interesting XML technologies really JUST beginning to come online now.

-- Kurt"