Re: Stored fields ordered left to right

Dawn M. Wolthuis <dwolt_at_tincat-group.com> schrieb:
> "Adrian Kubala" <adrian_at_sixfingeredman.net> wrote in message
>> You are not modeling with functions, you are modeling with lists. I can
>> tell because there was no mention of lists in the original preposition,
>> and lists are not required to describe functions, but nevertheless lists
>> have snuck into your model. On the other hand, the above would be a
>> perfectly good function-based model of "Person 12345 has the list (hope,
>> (cat, geneva, null), (dog, rugby, null))".
>
> My function is named PERSON and it maps the string "12345" to a set of
> strings or string tuples. I can use more precision in the future to ensure
> that you can see that this function provides a model for a plausible
> implementation.

I must have failed to communicate my point clearly, because I'm not contesting whether the function maps a string to a set, but whether it accurately models the predicate it claims to.

>> I say it's not modeling because I don't believe you have a general
>> theory for how prepositions (in this case) can be mapped to and from
>> lists in a general way, with a useful algebra on lists which
>> preserves the truth values of prepositions. It is not enough to
>> provide a post-hoc rationalization for why you chose these particular
>> lists for these particular examples. But if you do have such a theory
>> I am excited to hear it.
>
> That is one of the things I have been working on, starting with studying how
> the many developers who have worked with this model since 1965 have
> "post-hoc" been doing this and pulling out common, repeatable processes that
> are used.

I understand this kind of experience, and its lack is not what I'm complaining about. If you are programming in OO, it takes experience to decide which objects to include in your model. If you are creating a relational database, it takes experience to decide which predicates to include. In both these systems, once you have done that, there already exists a theory telling you what it MEANS to be an object or predicate, and how you can reason about them in a formal way.

> Although there is no written document on how to do this (to my
> knowledge) there also was no such specification for "secretaries" who
> set up filing systems in days before computers -- and yet the job gets
> done and the solutions seem to be quite flexible in meeting the needs
> of a company often for many years (many instances of > 20 years of
> such databases).

Just because something gets the job done doesn't mean it's a good model... the Babylonians, for example, could predict solar eclipses accurately but they did not understand that the earth orbits the sun and the moon the earth. Received on Thu Jan 15 2004 - 02:41:49 CET