Ellie [far37]

Francois-Rene Rideau rideau@clipper
Fri, 25 Jun 93 7:44:53 MET DST


  I've read the Ellie papers. The idea objects should group by interface
rather than by inheritance is good (that's what I did in my language,
even before I saw inheritance isn't needed at all; down with inheritance).
So the good idea is group objects by interface. The bad idea is that
interface contains only subnames exports. You can only say "there exist".
You can't say "there don't exist" or "that doesn't apply", etc. Only
strong programming conventions, and/or the use of the only negative keyword
"function" (i.e. does not depend on static values) can allow full negation
support in a clumsy Ellie sublanguage. I don't like the exports done only
by names either. When one can avoid names, better do so. What's important
is not the existence of objects (any object exists, or it's not an object :-),
but logical links between considered objects, that is how objects behave,
how the operate each on the other, etc.
  So, let's keep the interface aspect of the language. Let's keep the
grain-size adaptation aspect for a future release of MOOSE, and it's over.

Again, about C, let's use it only to boot the system. C is too low level.
C++ is either a dangerous low-level language or a clumsy high level language.
So I think we must build our language. After all, don't we have to build a
compiler ?

--
      ,						,           _ v    ~  ^
-- Fare -- rideau@clipper.ens.fr -- Francois-Rene Rideau -- +)ang-Vu Ban --
					'		    / .