MORE ON "MEANING IN DATA MANAGEMENT"

From: Don Mendelsohn

Date: 9 Dec 2005

The polarization of Darwen's more purely mechanical/how view/style against Pascal/Date's perhaps more human/why view/style with it's anthropomorphisms is a familiar one to me from other fields. In a way I'm straddling both the perspectives I inferred from the On Meaning in Data Management dialog. May I suggest a way to synthesize them, make terminology a little clearer (I hope) and make everyone more happy? But I won't say anything about teaching methods; I'm no teacher.

A telephone tranduces and transmits audio signals between people who wish to converse and understand each other. It would be hard for anyone in a developed country in this century to think/speak of the telephone or telephone system understanding anything it was transducing/transmitting. However, if the parties on both sides are to understand each other the system must posses a quality that, though it is mechanically implemented, is carefully tuned to the purpose of humans understanding each other (by engineers based on applicable theory of the field); fidelity.

May I suggest an analog to the concept of "fidelity" for data management - though you could come up with another word if you like (integrity is perhaps not general enough and anyway has it's own very well defined meaning in this field) - it's what's necessary for preserving the latent meaning of (I didn't say "in") the data to whatever level, for whatever purpose the DBA defines it should be preserved (mere archival/retrieval of bags to inferencing); this is the purpose of the constraints and relations of the DBMS. RM is the most general and comprehensive model we have, perhaps can conceive of for capabilities of DMBSs for preserving any sort/level of fidelity. This "fidelity" preserves whatever is necessary for humans to get meaning back out of the DB, but it is not itself "meaning" and the DBMS is not "understanding" any more than a telephone's transducers or companders understand human speech. If a human examines companding algorithms or hardware they may learn more about the frequency ranges most important for understanding of human speech (or limitations of transmission of the transduced signal), but that's because the engineers understood that, not because the telephone hardware understands it, in either a static or active sense.

When you see abundant examples of poor understanding and bad practice in data management, is it due to poor understanding of how meaning is latent in data, or poor understanding of the mechanisms for preserving the latent meaning for a given purpose, or both? Even of "both" is frequent, I still see them as somewhat distinct problems (even if attributable to less distinct causes; education, etc.)

From: Hugh Darwen

I have no comment to make because I find this kind of writing too difficult to follow (sorry, Don). I do get a feeling that I might be able to sympathize with the general drift.

From: Fabian Pascal

Re polarization: don't think so.

A TRDBMS is predicate logic and set theory applied to database management. It preserves as much of the semantics as logic can and a computer system can "understand" in the sense to which you allude.

Ed. Note:

… logic—is an analytical theory of the art of reasoning whose goal is to systematize and codify principles of valid reasoning. It has emerged from a study of the use of language in argument and persuasion and it is based on the identification and examination of those parts of language which are essential for these purposes. It is formal in the sense that it lacks reference to meaning. Thereby, it achieves versatility: it may be used to judge the correctness of a chain of reasoning (in particular, a “mathematical proof”) solely on the basis of the form (and not the content) of the sequence of statements which make up the chain. [emphasis added]

--Robert Stoll, SET THEORY AND LOGIC

where “lacks reference to meaning” refers to verb meaning.

From: Don Mendelson

You've both been more than kind in your replies to my muddled email. I'm going back to the logic (and other) books until I can express myself more clearly on technical ideas, and also differentiate them from more sentimental notions (I think the Alex Bunardzic quote that kicked off your dialog had both aspects) I'm grateful for the opportunity to communicate, yet feel apologetic for wasting your time.

Anthropomorphisms : "meaning" and "understanding" as applied to data and data management systems. However I like Mr. Pascal's qualification: "[TRDBMS] preserves as much of the semantics as logic can and a computer system can 'understand' in the sense to which you allude". I don't know if one of you refrains from using such terms and the other allows them only because of contrary ideas about their usefulness in teaching, because of contrary ideas about their actual applicability to machines (or even humans), or because of differences in writing style, or some combination. Perhaps "polarize" was too strong and premature a notion ... especially if you've come to some acceptable compromise with each other about the issue and how it should be expressed?

As I said above, I'll come back to this when I have the chops.

Posted 2/10/06