A few folks have had a side discussion about the idea of using the
new mixin mechanism to express kernel ideas like sprites. Among other
things, using mixins avoids the double-dispatch required when using a
function interface.
But this would mean that we abandon our previous position that the
kernel API should use just JS1-style functions and vars. So we need
to think through the implications. One important one is this:
- a pure-native kernel implementation (for example, a JME kernel
written in Java) would need to either double-dispatch under the
covers or else directly support the JS2 class system.
Either way, I think the presence of a significant question like this
about the structure of the kernel API suggests we should consider
labeling the kernel API "preliminary" for the OL4 release, with a
formal definition to follow in a subsequent 4.x release.
jim
