On Mon, Nov 25, 2019 at 8:12 AM VinÃcius dos Santos Oliveira via Boost
the task force interested in discovering a fundamental design.
I think there has been ample time for "discovering a fundamental design," since hasn't been a JSON library proposed to Boost for over 10 years. The premise of my library is simple: "No single JSON design can address all use-cases in terms of both functionality and performance," Therefore what I have done in Boost.JSON is to address a specific use-case and optimize it for that case. In particular my library is fantastic for network programs which prefer to interact with JSON through a DOM (the boost::json::value type) to leverage the benefits of ad-hoc prototyping. Perhaps there is a "fundamental design" remaining to be discovered, which is easy to use and does everything my library does, and also everything else (such as going directly to and from user-defined types), and also achieves the same level of performance. But I have not seen it, and from my early prototyping experiments I suspect that the goals conflict with each other and such a design is not possible, except for the trivial solution of having what is really two (or more) libraries in one. If there is evidence of this fundamental design please bring it to the list's attention. Thanks