Taxonomy of Facets & Composites

[Ken Kahn]

MarkM wrote:
>
> Good point.  In E's design (but not yet in E's implementation), we're very
> careful to be explicit about sources of non-determinism, and to reify them
> as capabilities, so that confined computation that is denied such
> capabilities is constrained to execute in a deterministic fashion.  If the
> inputs to such computation are logged and all its I/O is intercepted, the
> computation can be deterministically replayed. These guarantees hold even
> when the code was authored by someone that wishes to violate deterministic
> replay.  Computation that is constrained in this way can still wall-bang,
> but it cannot read the wall-banging of others.  Covert bits are prevented
> from leaking *in*.
>
So I/O includes message passing? Or is there some other way you deal with
the non-determinism of message arrival from multiple sources?

> Yes, I intend precisely for "objects" to be the quarks of our
computational
> physics -- much as Hewitt indented for "Actors".  That's why I explicitly
> introduced the notion of "composites" for descriptive aggregates of
objects,
> and sent that note proposing to start taxonomy of meaningful patterns of
> descriptive aggregation.  By my definitions, an aggregate of composites
*is*
> still a composite.  I believe "composite" is morally equivalent to
Hewitt's
> "configuration".
>
> So why don't I just say "object" instead of "composite"?  To emphasize the
> subjectivity of the decision about how to aggregate.  "object" suggest
> something that's really there, in some reductionist or objective sense of
> "really".  Perhaps I should just adopt Hewitt's terminology?
>

I'll agree that aggregation is subjective (though it might not seem that way
when a cannon ball is flying towards you but maybe when someone blows smoke
at you). But there is something dissatisfying about this picture. Consider:

Ontology 1. There are characters and text string that are made of up of
characters. Characters and strings are different types. You need ways to
combine characters to make strings, to add characters to strings and to
combine strings. And ways of breaking strings into characters and
substrings.

Ontology 2. There are only strings - some happen to have a length of 1. All
you need to combine and break them apart.

To me there is something very appealing about #2 - it is simple and does the
job. So why by analogy can't you build a system in which there are only
composites? Some just happen to be a composite of a single thing  that is a
composite of a single thing ad infinitum. Or least have a system where the
same "laws" apply to composites as atoms?

Also it seems kind of arbitrary how you bottom out. Is a character an object
or is it an aggregate of objects that correspond to a bit representation of
the character?

Best,

-ken