I think a version of Conway's Law applies. A technology succeeds if it mirrors the organization of the programming team/institution.
So an organization where there is a high-value managed data set and related guru/guardian class will use schemas: for example SQL relational databases. Or Document databases. SGML. XML Schemas.
But an organization where conformance of a transmission to a format is condidered a kind of unit test, will not use schemas. They will have used simple XML, and then moved on to JSON.
I have been making some web services with JAX-RS 2.0. It is all about removing the need to code the layer between object and MIME mefia type. I have been impressed at how irrelevant the features it provides are to services with fixed schemas or which use XSLT. So added to JSON's virtues we have to add the force that many data binding frameworks take to developer away from using Schemas and XSLT: it artificially makes some things hard.*
Greetings from Granada, Spain!
Rick
* An example from a couple of data binding frameworks: if you use POST they support MIME multipart form only. No support for ZIP, for example. Which is not required by http or mime or html (or even The JAX-RS spec).