Hi Avi,

> I liked this idea when you brought it up before, and I still 
> like it - it seems like a good Kernel Cleanup sub-project. 
> Andreas, if you've given this some thought, do you have any
> hints to those that might try to take it on?  What would be
> the Simplest Thing we could do to get started in this direction?

The Simplest and Most Important Thing to do is to figure out how any new
scheme can be used in parallel to the existing source code management. As I
have learned (the hard way; and I'm pretty sure Craig will agree on this -
if not today than surely later ;) trying to change these kinds of things is
a real pain if it means you have to find and fix all of those implicit
assumptions in the system while you're trying to come up with something
better.

So what I'd do is to look at SourceFilesArray and friends and see if you can
tweak out a few bits for representing an (initial version of) that
MethodHome. I'd probably look at halfing the range of either changes or
sources file (or both) to get to that initial set of bits. Once you got
this, you have freedom to play with everything you want.

The next thing I'd do is to look at how the current sources and changes
could be represented to fit into the bits range that you just got yourself.
This ensures two things: For one thing it means that once you've succeeded
you can painlessly convert the old into any "new" representation and test if
the system actually plays well with that encoding. Secondly, it means you'll
be thinking about how to compactly encode the sources+changes, both of which
is very important on smaller machines.

Then you're basically free to go into any direction you want.

Cheers,
  - Andreas