Hello,I think the idea is good, personally, I use lua a lot, and really like how it's done, but it lacks all the "modern" stuff you mentioned (threads, net, graphics...), I would highly recommend to study it closely (lexical scope, closures...).
About the implementation, I would love to see it in C instead of C++, or at least with a pure ANSI C API. (damn, I HATE C++, but sadly I know it has tons of good reasons to be here and be used, damn... we are producing CO2...)
And about the language itself, what do you have in mind? Dynamic? In a VM? Compiled?
For the unicode thing, I think it's time to handle it properly, many many MANY, piece of software claim Unicode support without even knowing what it means. Calling "is_number()" on 七〇〇 (700 with japanese kanji, often used in restaurant to show the price), should definitely returns true, and calling "to_number()" on that string should return 700. A lot of work has been done, everything is here (lookup tables, conversion tables...) waiting to be used. I know nobody care (or?) about the feature I just mentioned, but that's for the example, what I mean is that if you start something new, make it really good, not just "cool".
It pisses me off to have to write tons of code all over again every time I want an OpenGL context, and please, don't answer SDL or GLUT or Allegro, I want a native type with natural use, I want to move my context between threads without wondering, while being sure it use proper CGL on MacOS X and proper whatever on Linux and Windows to make it fast.
For this, I also think it's time to select a few platforms. I would go for something like Windows, MacOS X and Linux. The key point is being that the language should not be limited because platform X, used by 2 folks, doesn't support feature Y. Having a proper module based architecture would bring the core language to a lot of platforms without effort, but because of some core features, some platforms have to be removed.
Voilà, that's my 2 cents, best of luck... ho, and if I may help, poke me, but hell, I'm not touching C++, after all those years, I'm finally out of it, so no, I'm not diving again, not even for the greatest breakthrough of all times!
-- Nicolas Goy, Programmer Goyman.com"It's easy to write a function, the real work is to give it the right name."