Well, having UIMLs/JSON to describe UIs is no doubt a very neat option, but it'd take away the initial notion of an extremely layered down Shoes-esque way of interface building and functionality (atleast for Ruby). In my opinion, wrapping the UI part of the Haiku API is the most challenging part for any bindings. If there is a powerful declarative UI mechanism (in JSON or XML) and there is a C++ parser that constructs the UIs from that, you can wrap that library both in Python and Ruby then you have a common declarative UI and you can implement the rest of the app in a sort of View-Controller(-Model). That seems like a nice idea . But, writing up a parser/library in C++ would obviously take some time + the language specific port. Or we could just use the safe way and make the language specific parser + generator similar to how RubyQT works, generating (rubyuic) --> class Interface files --> includable in ruby/python files to separate implementation and the interface. But of course defining the declarative UI format/mechanism should have its own share of discussions? I don't think the above would apply, since these are two different projects with some common goals. One is for making Haiku bindings for Python, the other for Ruby. Though there are a lot of commonalities and I suppose both Ankur and Akshay may want to realize that Haiku probably won't be able to sponsor both with a limited number of GSoC spots. Yes, I guess we totally understand that. -- 4, 8, 15, 16, 23, 42