As part of our web application framework (currently being ported from Windows to Linux) we have a JSON parser and library that might be of interest to a wider community as part of the Boost libraries. It uses a Spirit parser that deals correctly with embedded UTF-16 inside strings. The JSON blobs themselves are built on Boost.Variant and can store null, booleans, strings, int64_t, doubles and arrays & objects. The strings are our Unicode string implementation -- how this should be dealt with in Boost is an open question maybe only cleanly soluble when Boost has it's own proper Unicode string handling. A templated adapter layer could be used I think for both the string and the numeric types. There is also the internal implementation of the array and objects which could be moved to a policy based design. At the moment we use std::vector and std::map with boost::shared_ptr< json > as the value type. The JSON objects are immutable and use a thread safe CoW implementation when changes are required (handled through a separate jcursor object). One thing that is missing is a good EDSL for creating JSON blobs within C++ source code. The latest implementations is where we are using it to create an ACID JSON database and is here: http://svn.felspar.com/public/fost-base/branches/dbschema/Cpp/include/fost/j... The parser itself is here: http://svn.felspar.com/public/fost-base/branches/dbschema/Cpp/include/fost/j... There's a lot of room for improvement, but hopefully it's a good start. K