Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

SUO: Re: Missing Ingredients




o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o

MI.  Note 9

o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o

TJ = Tom Johnston

TJ: You have my vote for stepwise refinement.

Hah!  Gotcha!  Stepwise refinement is a top-down method!  What day is it!?
What difference does it make on this Moebius Trip!?  At any rate, beyond a
certain point of the "analysis by synthesis" (ABS) rotary brake on all your
average logico-gravitational analogies, we never really have much sense about
which way is Up any more, and we have to fly by the seat of our pants, as all
the old pilots used to say.  And yet, locally and relatively speaking, it still
makes sense to keep on working from lower to higher and from higher to lower in
some moderate sense of those terms.  I think that this is something to think on.

TJ: As for reliance on mother tongues, if this is an admonition to
    fiddle as little as possible with the ontologies implicit in the
    unreflective linguistic use of the SMEs (subject matter experts)
    that I deal with so often, then I differ.

Not sure.  I was merely trying to handle the tangent that Rich was
flying off on, apparently with no more stimulus than my use of the
word "sentence", which I never imagined would be taken as anything
more "natural" than the formal logical kind of a sentence, as that
is all that we can really have much hope of handling any time soon.
I say this as someone who kept up pretty well with the research on
various flavors of linguistics, natural and unnatural, for several
decades, up till just recently, when I got busy with other matters.

TJ: I find that my fellow constructors of data models
    usually do fiddle as little as possible with what
    the SMEs tell them, and I think the data models
    based on that approach aren't as good as mine,
    which are based instead on an ontology which
    arises from dialogue in which I question their
    assumptions, and suggest alternatives.  Almost
    any bad data model can be made to work, i.e. can
    be part of an information system which meets stated
    requirements.  It just requires lots of otherwise
    unnecessary code to compensate for the lousy set
    of schemas.

Yes, and there are fundamental information-theoretic reasons for this.

But is this level of vague generalities what you meant by bottom-up?

If so, I must be on the wrong plane.

Jon Awbrey

o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o~~~~~~~~~o