Sectionization

[David Chapman <Zvona at MIT-AI>]

    I think all of this can be short-circuited if modified sections get re-sectionized
    whenever you need the sections to be consistent.  

That is what I do now (using check-interval-sections.)  The point
is that that isn't as good as one would like; some of the
information you want isn't there.  Just some of the lossage modes
available:

-- A single section can have any number of functions in it.
There is no way of knowing which have actually changed.

-- New text may go in as inferior to the buffer directly, not of
any section, even if the buffer has been sectionized.

-- The previous and next pointers are not at all trustworthy. 

-- Sections that have been deleted from the text altogether may
persist as inferiors of the buffer.  (This is probably an actual
bug, unlike some of these others, which are non-features.)

In general the sectionization structure just isn't reliable
enough to support the sorts of things that you would like to
build on top of it.  Now my guess is that the (many!) bugs in
Compile Changed Functions could be patched up, because it doesn't
matter to much if you compile a few extra functions.  But it
might even be easier to do the right thing and incrementally
resectionize.