Re: Extending my question. Was: The relational model and relational

From: Steve Kass <>
Date: Thu, 20 Feb 2003 05:00:40 -0500
Message-ID: <b328t6$5s6$>


  First of all, we clearly disagree on what makes a model useful, as well as on boundaries. You separate things into logical, physical and conceptual. I think we're pretty close on what the physical is (implementation), but it ends
there. My logical model is the mathematical one, and I don't require it to have suitability for machine processing (I imagine you see my logic that way). I require it to be consistent and complete, and I value logical models
that do translate to machines well. My third model is the real world.

I'm not astounded by the fact that you see things differently, I should say, but I do find it unfortunate that you don't seem open to the possibility that
there is more than one way to look at things.

You and Date insist that cardinality can only be determined by counting, which
requires distinguishability. I have a scale that allows me to determine the cardinality
of a multiset of tuna cans by their collective weight. I also know how to count,
but sometimes it's quicker to use the scale, and it doesn't require distinguishability.
If you're happier when there's only one way, you've certainly made the right choice. I misplace my scale far more often than I misplace my fingers.

I also guess we'll never settle the case of Date's screwy example. Your claim
that just because I like multisets I should be happy for anything to be duplicated
is just silly. Believe it or not, I think there are reasons why some tables work well
with duplicates and others do not. A supplier-parts table should not have
duplicate rows, for reasons that are clear to me. You also agree that such a
table should have distinct rows, but somehow you believe I'm obligated to make sense out of duplicates because I find sense in duplicates elsewhere, when
you say "

If you are a proponent of multisets, then there shouldn't be anything wrong with allowing both to be bags either."

For the record, I'm a proponent of ketchup, too. Please dump some in my coffee, since there shouldn't be anything wrong with that. With any luck, Ill object, and you can patent the brilliant proof of the once-elusive theorem that ketchup serves no useful purpose.

I'm not at all ashamed of the fact that I like ketchup, on some things, or that
I value multisets in some places and recognize their inappropriateness elsewhere. Sure, sets always provide a solution. You can make them work everywhere (if not surprisingly, since all mathematics reduces to sets).

I think other things work even better for some purposes, that's all, but you or Date or anyone else are welcome to stick to the tried and true, and it won't offend me in the least. You can use Roman numerals and shells, for all I care. God knows they have a much better track record than any piece of software does. But as you've probably guessed by now, I'm stubborn, and whether down the road I decide that multisets are a really bad idea or not, I doubt I'll have any regrets about what I learned.


Received on Thu Feb 20 2003 - 11:00:40 CET

Original text of this message