Re: A pk is *both* a physical and a logical object.

Cimode wrote:
> On 13 juil, 17:47, paul c <toledobythe..._at_oohay.ac> wrote:
>

>>paul c wrote:
>>
>>..
>>
>>
>>>"class-concepts".  OTOH, as far as he was concerned the word "men" stood
>>>for all people!
>>
>>Now that I think again I might have been wrong to attribute that to him,
>>  I now seem to remember him writing that "all men" mapped one-for-one
>>with "all rational beings".
>>
>>p

That seems a pretty open-ended question to me. When it comes to math models, I have experience as strong evidence that most mathematicians would consider any model of mine puerile. I mentioned Russell because his attention to grammar, at least when it comes to relations as they concern computers has always seemed to me obsessive, way too fastidious as far as the usual commercial application is concerned, but Russell was writing for a different purpose and maybe I shouldn't blame him for dying a few years before Codd's first papers were written. It reminds me of many earnest people I have known who were fastened to ER and other kinds of external modelling philosophies. I usually thought they were wasting too much time but I also noticed that almost uniformly, they were oblivious to my opinions that they were dilly-dallying and some of them even expressed polite pity for my ignorance. For me, many of the subtleties and nuances that resulted from their models had nothing to do with the business problems at hand. But I have to get along on the same planet with those who have that attitude, so it seemed polite to say that they might be on the same page as Bertrand R., who wrote so much about relations. When you implement relations on a typical machine, the abstraction is such that it only makes sense to give up many nuances and stop nattering on about them.

That might sound like so much editorial. Establishing a connection between an "identifier" and reality has never bothered me and at my age, I doubt it ever will. That may be because I've never done any scientific applications but when it comes to business, all it needs is for the client to accept it and agree to use it. I my name is is in a system along with some "ID", I know as a customer that some business operative may refer to me as "him" or "Dear Sir", but for me that is just natural politeness, nothing to do with the system. A relation may start with subjects and verbs but once it is implemented, those terms become moot in a physical machine and all that matters where identity is concerned are attributes and uniform treatment of them. That is the nature of digital machines. Users must agree to interpret such uniformly, which is not a computer issue.

I think Codd's Information Principle is paramount, more important than which algebra is used but equally paramount with what is operable, ie., practical, given the available and affordable machinery. I remember using a 17-byte character identifier, at least accomodating POSIX, for customers who wanted ascending identifiers, just a time-stamp to the nearest one-thousandth of a second and getting the client to agree that their application would not need to commit more than 1,000 transactions per second. For that matter, the Information Principle seems to naturally apply to far more than materialized relations/tables and it amazes me that I never seem to read of it being applied by transaction theory writers.

I'll shut up for now.

p Received on Sat Jul 14 2007 - 03:18:19 CEST