New Model structure

[Eric Dean Tribble <tribble>]

That's ok.  I wouldn't think saying it's a set would obligate a
   duplication check.  It would simply allow it.  
   Remember, {a, b} === {a, a, b}.

No.  When I remove the element, I have to examine every element.

      Consider the Set version of this:  If I add a dependent twice, then
      remove it once, should the dependent still receive updates?  

   Hmm.  Interesting.  Note that this is also a possible reason for MuBag
   (in addition to MuSet).

Mutable vs. immutable is only a matter of implementation speed for
dependents.  There is guaranteed only one reference to the dependents collection.

Dependents can partially represent state by the set of models on which
they depend.  The following example roughly describes something in FM.
The selection object for a header in an outline only depends on the
selection object for its text selection if the selection falls
entirely within that header.  Otherwise the entire header is selected,
regardless of the cursor position within the header.  Hmm.  I'll draw
a picture later.  I'm not sure this is currently implemented using the
dependency mechanism....

   Huh?  What about "model->dependents()"?  

There's a funky optimization which I'll describe in person.  It uses a
special collection class.  Ask me about it.

End of thread.
So there :-)
dean