What's the most optimal format?

[mohammed90]

Data formats are a dime a dozen. The hope is to pick a format that has good tooling in almost all programming languages and environments to facilitate testing everywhere (as possible). The format is preferrable to have a schema variant so invalid/unexpected data cannot sneak in.

Studying prior art of projects with similar aspirations we have the following:

• cache-tests: The project uses JavaScript, allowing it to facilitate browser and server-cache testing. It was reported that some users hoped to use the same test suite for their cache implementation in their language of choice, and they had to option except to re-implement them in their language.
• structured-field-tests: The test cases are defined as JSON, which has great portability but has some drawbacks. It doesn't support comments, and I remember experiencing pain with large numbers (I cannot remember the details).

I considered and discussed Protocol Buffers because it comes with schema definition and a textual format, making it possible to write the test scenarios by hand and perhaps be partially machine generated. Protocol Buffers support isn't mainstream for many languages. There's official support for 7 programming languages, and the rest are community maintained if available. Thus the protocol buffers consideration make them less portable than JSON.

What other data format are to be considered? Is JSON the strongest contender?