reflective architecture

Ken Evitt cpe2@gte.net
Mon, 24 May 1999 22:56:15 -0400


If a computing system was composed of fundamentally uniform entities that
are manipulated in a consistent manner and the system contains entities that
represent itself--then what else is necesary for the system to be
reflective? In other words, if a computing system has unified system
abstractions (with abstractions of itself) then is the system reflective? My
question concerns the implementation of the TUNES functionality. Is the fact
that TUNES is objects all the way down and that some objects represent the
system itself enough of a reflective architecture to allow "anything to be
programmed that could possibly be programmed"?

I am wondering because it seems to me that Retro is where things are
happening. What I see as the means to TUNES is to implement it at not a low
level, but at a low layer, meaning that TUNES should be the fundamental
layer of abstraction and everything should be an extension of it. Can
Brian's Arrow System be implemented within Retro or something similiar? And
if so then could we just work within the realm of the Arrow system from then
on? What is it that other people are wanting to do? Do you want to implement
the HLL on top of something? I don't really see how that would work because
I don't really see that the HLL could every be on top of the operating
system--it IS the operating system. Or am I wrong?

Ken Evitt

P.S. Brian, what does your Smalltalk code do (and how do I do it)?