[
Lists Home |
Date Index |
Thread Index
]
On Tue, 7 Jan 2003, Roger L. Costello wrote:
> Now let's get back to the hard issues:
>
> - should there be 2 schemas, one for fundamental data and one
> for derived data? I will argue that there should only be
> one schema - the fundamental data schema. Derived data is
> transient and should not have its own schema. What do you
> think?
>
> - at what point does sharing of fundamental data become a
> Service of derived data?
I honestly don't think you can always make such a distinction. SUppose you
have a set of fundamental data x, and derived data y, with the monotonic
relationship
y = f(x),
Then a simple transformation can make y fundamental, and x derived
x = g(y) (g = the inverse function to f)
So it would seem you need something to define the semantics of either
data model (so that interpretation is consistent), but can't really say
which one is 'fundamental'.
Consider your example: which is more fundamental, cartesian coordinates
(x,y,z) spherical (r, theta, phy) or cylindrical (r, theta, z). The answer
is -- all of them.
Ian
|
