Thinking about binary diff and merge

[Martin Pool]

I like the way you're thinking about this.

I think ideally we would match up through a 'category' abstraction,
which can be determined by globs in .bzrrules or possible by something
else in future.  I think that is more flexible than specifying it has
to be a glob or has to be a mime type.  You can have different files
all of the same mime type that should be handled differently.

As much as possible it would be good to make the definition of how to
run different diff merge things work without writing new Python, so
it's more accessible to users.


-- 
Martin