Hi Gregor:

>From gregor@parc.xerox.com Mon Nov 17 11:28:05 1997
>To: lieber@ccs.neu.edu
>CC: lopes@parc.xerox.com, dem@ccs.neu.edu
>
>Clearly the term "meta" is relative.  But I've found that it works in
>practice because one so rarely needs to deal with meta-meta-objects.
>So, one gets by with just the three levels: objects, metaobjects
>(including classes) and metaclasses.
>
I agree with you 100%. However the UML and OMG people think 4 layers,
including mete-meta are better.
>
>If I understand your picture I agree it captures additional
>distinctions, but I'm not sure when I need them.
>
You use it when you work with two tilings 
of the Interpreter pattern described in GOF, as described in the 
OOPSLA '97 paper by Lorenz.

-- Karl