Erann:

Here is my current view of Smyth's Law of Demeter and its relationship
to the original LoD:

The LoD is well defined in our papers. There are different versions,
like the object and class form and the strong and weak form.

A good short summary of the weak object form is (quoting Smyth):

>   A method shall act only upon the arguments and the pre-existing
>   state of the object.

The LoD can be applied to many different languages: There is a 
LoD-CLOS, LoD-Java, etc. We can also apply the LoD to 
multi-threaded languages
without return types and this combination seems to be
very useful for distributed/parallel/multi-threaded systems.
The application of the LoD to 
multi-threaded languages without return types I call
Smyth's Law of Demeter.

Does this clarify the situation?

-- Karl

>From gat@aig.jpl.nasa.gov Mon Jun  2 13:22:47 1997
>Date: Mon, 2 Jun 1997 10:22:30 -0700 (PDT)
>From: gat@aig.jpl.nasa.gov (Erann Gat)
>To: lieber@ccs.neu.edu
>Subject: Re: LoD and return values
>
>
>> You are right:
>> 
>> "essential" was not the right word below.
>> 
>> "convenient" would be better.
>
>This still doesn't resolve the apparent conflict.  Smyth's position is
>that returns values are the moral equivalent of a goto -- something to
>be avoided at all costs.  Your modified position, that return values are
>"convenient", still seems at odds with this.  On your view, return
>values are acceptable in good code; on Smyth's view they are not.  So
>there still seems to be disagreement, and I'm having difficulty reconciling
>this with your statement that Smyth's formulation is a "significant
>improvement."
>
>> You can always avoid return values by accumulating the result in
>> an argument.
>
>Yes, that's another issue I was going to bring up.  Return values are
>really just a syntactic convenience.  The formal properties of a language
>without return values are identical to those of a language with return
>values, since one can emulate return values in a number of ways.  (Passing
>continuations around is my pet technique.)  But this is an orthogonal
>issue.  I'm just trying to get clarity on whether there is agreement
>about what the software engineering principle called the Law of Demeter
>*says*.  That seems to be a prerequisite to any cogent discussion of its
>effects.
>
>E.
>