The second section discusses SOAP/REST and XML/JSON, saying that XML Schema attempted to solve two problems (document formats + object wire format), but it predictably became too complex. He claims that we now use REST (not SOAP) and JSON (not XML), so the "impedance mismatch" problem is now solved.

Is this true?

JSON is used mainly where there is Javascript (ie. client side webpages - though that is becoming more and more inclusive these days... almost everything is a webapp (exceptions?)) - but it's schemaless (or typeless). While this is popular (eg key-value NoSQL stores, and dynamic typing in Ruby, Python, Perl), it's not right for everything. If you did want typed data, you'd probably just use XML Schema, since everything is already there.

I ask because of a startup idea based on XML Schema. It could survive XML Schema going away, but I'm not so sure about the apparent long-term trend away from schemas/types altogether. With fast-enough computers to nullify the speed advantage of static types, will the remaining advantage of type-safety be valuable enough on its own - What is the evidence for and against? Even databases are going schemaless (it seems).