Shannon revisited

["Vegt, Jan" <Jan.Vegt@softwareag.com>]

Len,

Coming back to our earlier discussion I found out - thanks to Gilula - that
there is an obvious and important case where Shannon is relevant when
talking about data(base) schemes.
When the scheme is seen as an abstraction of reality [ mapping neuro-logical
to binary space ]it is obvious Shannon's coding principles apply. Thus
signal/noise, entropy and related measures apply. 
When you look at the scheme as an abstraction of the data inside [
digital-digital domain ; or the 'inside' view ] Kolmogorov is useful.

I believe this is relevant for several ongoing discussions. Let me quote
Gilula (1994) on the relational data model : "Underlying all these
suggestions, however, there is a tacit but important assumption - an
assumption that can be a rich source of confusion if not properly understood
and appreciated. That assumption which I will refer to as the interpretation
principle, is as follows : The data model in question must have a commonly
accepted (and useful) INTERPRETATION; that is, its objects, integrity rules,
and operators must have some generally accepted correspondence to phenomena
in the real world."

He goes on to claim that because of our inconsistent interpretation of
reality  the relational data model has fundamental flaws. Oops, that's a big
claim when read here in isolation. Let's leave it there for now. (see my
reply on XML and RDBMSs)

Now read this on type and data abstraction (Cardelli and Wegner, 1984):
"When looking at a piece of raw memory, we have generally no way of telling
what is being represented. The meaning of a piece of memory is critically
determined by an external interpretation of its content.
This universe is untyped because everything ultimately has to be represented
as bit strings: characters, numbers, pointers, structured data, programs,
etc.
"Untyped" actually means that there is only one type, and here the only type
is the memory word, which is a bit string of fixed size. Untyped universes
of computational objects decompose naturally into subsets with uniform
behavior. Sets of objects with uniform behavior may be named and referred to
as types."

Clearly this deals exclusively with the digital domain but I think we can
safely conclude that interpretation is important. Obvious I know.

Communication is based on agreeing on established meaning, or consistent
interpretation. But, innovation can be defined as deviation from established
meaning. Human communication is about progress.

To good not to mention : the motto of Gilula's book : "The reasonable man
adapts himself to the world: the unreasonable one persists in trying to
adapt the world to himself. Therefore all progress depends on the
unreasonable man." (Bernard Shaw).

Jan

Disclaimer : my views are my own, not my employer's, unless proven
succesful.