=========== message of June 24, 1997
What is the meaning of 
a1#A . a2#A . a3#A

For the class graph 
A = B.
B = [A].

we get the traversal graph:
A1 = B1.
B1 = [A1] [A2].
A2 = B2.
B2 = [A2].

We would like to write a visitor method:
before a3#A to indicate that code needs to be executed on the
third visit of an A-object. Therefore, visitor methods
should be able to refer to strategy nodes.


For the class graph
A = .
the traversal graph is empty.

The conclusion is that the meaning of "from A to A"
is empty if the class graph does not contain a cycle
containing A.


========= Boaz message before semantics simplification

From boaz@ccs.neu.edu Wed Jun 18 23:36:51 1997
To: lieber@ccs.neu.edu
Subject: Re:  from A to A
Status: R

If we are to accept trivial paths, then the current decomposition
rule dictates that a strategy graph edge A->A must be interpreted 
as the trivial path <A> (that's because of the "to-stop" decomposition).
But that's not really necessary: There's an easy fix which allows
all cycles through A, and relaxing the semantics to require that
"there exists a decomposition..." instead of specifying precisely
how to break the paths.
Boaz