Information Microstructure, Key to Simplicity

                    Ed Lowry     
                    Advanced Information Microstructures
                    14 Old Village Road
                    Acton Mass 01720
                            508 263-3508
                            eslowry@mcimail.com

What is a reasonable structure for a data object, and why? Anyone
who designs information systems without a sound answer to
that question risks producing systems which are unreasonable in
much the same way that square wheels are unreasonable.
Foundational knowledge in most fields of science and technology
consists largely of knowledge about fine structure of the subject
matter. In information technology, understanding of how the
quality of information is affected by the fine structure of its
representation has been neglected. One result has been pervasive
excess complexity, but it can be corrected. Designing a formal
language for maximum simplicity of expression leads to the
simplest kind of data object, the unlabeled directed arc or
simple pointer. It can be shown that:
     For sufficiently large deterministic languages of a given
     size, those which provide maximum simplicity of expression
     across any sufficiently evolving set of applications must
     use data objects which are unlabeled directed arcs
     exclusively.

Complex data objects can be decomposed to produce additional
objects representing useful abstractions which contribute to
overall simplicity. The conclusion applies broadly to technical
descriptions prepared either for people or machines. Arcs are
theoretically optimum for large languages and rich applications
but empirically optimum almost everywhere. 

There are about 25 similar cases where optimizing an engineering
value leads to a simple irreducible structure rather than a
tradeoff. Those include: round wheels, tubular pipes, binary
memory elements, vertical pillars, etc. Merging functions on
directed arcs for many application domains into one language can
eliminate restrictions on the generality of formal language
semantics.  

Knowledge workers succeed by working CAREFULLY with various kinds
of information. Learning to do so pervades education. Careful
work of almost any kind requires understanding and control of
fine structures.

It is proposed that development of the Internet and any ongoing
information system be designed using arcs to model the whole
system and the information in it including legacy structures to
maximize simplicity and ease of large scale integration.