Funny.

RFC 4627 explicitly says this in its introduction, of course, but that is probably best treated as non-normative (4627 is not a specification, in any case). The detailed production rules confirm it though.

I came across your blog post because I had a client who insisted that JSON wasn’t a subset of Javascript, and hunting around the source of this urban myth seemed to be your blog. Most other sources (including the JSON site, wikipedia, etc) correctly state that JSON is a subset of Javascript.

In the context of something like:

var foo = {…};

your example is perfectly valid and legal Javascript, it conforms to the ECMA-262 (BNF in section 11.1.5), and it runs on the major JS platforms: I just tested in SpiderMonkey, Rhino, v8, and Futhark, all accept it fine and allow me to manipulate the elements in the object after initialization.

{ “411″: “yohoho.com”, “xmlns:x”: “Invalid character”, “Company name”: “Spaces”,”": “Zero-length string” }\

Valid JSON. Invalid JavaScript.

You know that Javascript allows quoted keys for its objects too, right? So this is fine in javascript:

var foo = {“new”:1, “var”: 2, “break”:3};

The only thing about this format is that you can’t use Javascript’s syntactic sugar of property access, but the Javascript is valid and you can still use it fine.

RFC 4627 explicitly says property names in JSON must be strings. It also says that JSON is a subset of Javascript.

“In fact, JSON objects can be invalid as JavaScript objects.”

Can you give an example of a piece of valid JSON that wouldn’t be valid Javascript?