Heterogeneity

[LANDAU_CHARLES@Tandem.COM]

I think there is still some confusion between mappings and data

A composition of two data objects does not create new data. It is really
just a mapping of the data objects.

You can distribute data objects, but I claim it doesn't make sense to
say you distribute a composition of data objects. If I have a
composition of data objects A and B, it could be the case that someone
else has direct access to A. The distribution mechanism can't work
properly if I only give it the composition, because it can't prevent
access to A.

In order to distribute a composition, you would have to say that you
distribute the underlying data objects, and create a corresponding
composition on the other machine. But since page sizes vary, you won't
always be able to construct a corresponding composition that is
completely accurate.

This problem is similar to the following problem, which doesn't involve
composition: Machine X has page size 4K and data object A which is of
size 4K. Machine Y has page size 16K. On Y, you will be able to write up
to 16K.

I think the answer to both problems has to be, if your machine has an
inadequate page size to accurately represent an object or composition,
we will do our best to give you access, but we can't prevent reading or
writing data that doesn't really exist, and therefore, we can't prevent
race conditions and we can't catch all the bugs we could catch if the
page size were smaller.

I think a distribution mechanism could always create a corresponding
composition that obeys that rule.