Parts should know their whole

Emptist wrote:
> Hi,
>
> I had a post related to the topic, see
> http://groups.google.com/group/comp....ow+whole&fwc=2
>
> Now after thoughts, I believe parts should know their whole, which
> object they "belong to" (one to one or many to one relation).

Three cases:

1. "part"s of an integral "whole"

a. initial construction from parts results in an immutable whole.
b. removing any such part destroys/invalidates the whole.

2. "part"s is "part"s

a. the "whole" is merely short hand
for "all of the parts" of a group,
and adding/removing/replacing parts has no effect
on the meaning of this short hand.
b. such parts might change over the lifetime of the whole.

3. "part"s of an abstract "whole"

a. any given "part" may be contributing to the realization
of many different "whole" things at the same time.
b. such parts might change over the lifetime of the whole.

The first case has properties which suggest
that it might be easy to benefit if the parts
know the whole - because the relationship,
once formed, is static.

The second case implies relationships which change,
thus requiring updates and synchronization. Still -
the parts only need to know the one whole that exists
at any given moment - it might be worth the effort -
depends what the parts need to do.

The third case adds the complication that each part
would need to track multiply many changing "whole"s
just to stay current. It is not easy to imagine a
situation which would benefit if these parts
know their wholes from the assemblages grown.


Cheers,

-cstb