yes, I do agree. JSON is not new any more and many good solutions exists even in Boost. I'd rather see something where you can define tags and their structure and have a compile-time solution providing the smallest parser possible. Maybe something like a string data packets parsing library would be great too, where you could either define tags in text format or simply bit fields or byte fields. I'm dreaming a little bit. So for any "not-too-complex" format, you'll be able to define a simple parser with functions being triggered any time you meet a tag or recognize a field. Boost.function, boost.signals2 etc would help. Maybe a Boost.Streamdecode library would be great. It's a standard problem. Or a Boost.ParseTags library... My 2 cents, David