here's as far we have got.

Folks,
        I should say that what Gilles pointed out is sound,  and,
        except  for the description of the 'holes', I do like the
        idea to simplify  the  registry.  I  do  not  agree  with
        Marten  and  Tony  when you say that a component is never
        routed,  or  we  have  to  change   the   definition   of
        'component'.  Components (except for holes) bring nothing
        in case of simple aggregation, and in case of proxy,  the
        components  HAVE to be routed.  Could anybody provides an
        example (except holes once again), where a  component  is
        really  needed  i.e.  where  the  component  shouldn't be
        registered as a route ?  I don't see any (but  I  may  be
        wrong :-)

        The big pro if this proposal is simplicity.  We just have
        to  find  a way to represent holes, but this can be a new
        field.

        Concerning Jessica's concern on the interas-out metric, I
        do understand Merit's point of view, and it cannot really
        be implemented as a local field since it's something  the
        tools  will need when tracing inside 690...  However, I'm
        not still convince it fits very  well  with  the  current
        scheme...  I'm puzzled.  The point is  that  we  HAVE  to
        find a solution!

-- 
Jean-Michel



-------- Logged at  Mon May 30 12:08:05 MET DST 1994  ---------