Re: Little design mistakes that can be easily avoided (2): Listenning to CELKO (and CELKO alikes)
>> Regarding the list above, I only think that it is the rare database that can tolerate multiple perceptions or multiple realities. <<
LOL! The advantage of mono-theism -- the universe makes sense!
>> Lots of db's are subjected to those, but I would say they usually don't tolerate them very well. To be a practical db, I think it's implicit its context is one common agreed understanding or interpretation among its users. <<
Let me give you an example. The Chinese are an ancestor worship
culture. Western Civilization is not. The concept of "Your father's
oldest brother's second oldest son" is "cousin" in English and a
special word in Chinese. With obligations to them.
The place you see categories making a difference is in OLAP and DW.
My favorite story is a consulting job I did with a shoe company. The
manufacturing side of house saw "steel toed work boots" as one
category. The marketing side of the house saw "LARGE SIZE steel toed
work boots" as things to sell to construction workers and "SMALL SIZE
steel toed work boots" as things to sell to Goth Girls, without no
real market for for the US 8-12 size range.
>> Taxonomies are useful for organizing discussion but dangerous when followed naively. <<
Agreed
>> Celko is probably abusing them in order to sell to simpletons or wishful thinkers or the habitually lazy, which seems to be either his chosen market or one that he can't avoid and vice versa. <<
no, to start discussion; you said that already. Duh! Unfortunately,
you are right about unavoidable market. More and more application
programmers are being told to do RDBMS and not given any training or
theory or even a book!
>> I don't think there is such a thing as a physical key. <<
Stand in a physical position. Does it uniquely locate you? YES! Can
I grab you and declare "He is here, Officers!!" as I turn you over to
the police? Yep!
Now I might want to have a (longitude, latitude, elevation, timestamp)
triple for your exact location in time and space or I can be a bit
vague (he was blind drunk and face down on the floor of the Pink Pussy
strip joint on the evening of 2007-05-30).
The co-ordinates are a trusted encoding with external verification
(GPS). Like wise the timestamp is a trusted encoding with external
verification (UTC).
>> Usually I think people are talking about indexes when they use that phrase. As much as anything it's the fault of old-timers (my generation) whose understanding of Codd's ideas was crude at best or more often completely ignorant. <<
Amen. I got it earlier than most because of two separate Masters in
Math and Comp Sci. But there was a learning curve in the early
days.
>> Lots of times, a generated ID is useful when nothing better is needed or when nothing better can be found. <<
In the world of Google, DIN, ISO, ANSI, etc. there is a Standard
99.90% of the time. What bothers me is that when a local identifier
is needed, nobody thinks about the design of the data encoding.
Validation. verification and risk of error are ignored.
>> Some historical writers are certain that there was a Soviet agent number nineteen in the WWII American White House. That is the best ID for him or her that anybody has come up with so far. <<
The NULL agent? :)
Received on Sun May 27 2007 - 22:41:43 CEST
Original text of this message