A nice idea is to fork AntiGrain 2.4 into something Boost-like and integrate it with GIL. If I had time to spare, this is what I would do.
Having worked with GIL now for over a year, I can honestly say that it is a real break through for graphics developers. The concept of a submimage_view is very supportive of creating generic functional components, similar in a way to your spirit library. GIL is a superb base upon which to build a graphics library. Working with experts in functional programming like yourself, using GIL as a base, developing a graphics library for simple primitves, such as lines, circles and polygons would be a trivial matter, and its already basically finished. Anti-grain is a half-finished library which works well in areas. However, I'm persuaded that GIL is the library upon which I think will ultmiately prevail. With adobe supporting the GIL library, who support functional programmers with many of their open-source offerings, it seems like a much better bet to go with adobe. I've moved my application to GIL and have not looked back. I encourage you and any other graphics programmers to give it a strong look. If I can persuade you to look harder at GIL, I would love to pick your brain about how to make a graphics library that works as well as your spirit library.