Re: 3 value logic. Why is SQL so special?
From: Volker Hetzer <firstname.lastname_at_ieee.org>
Date: Thu, 21 Sep 2006 23:26:18 +0200
Message-ID: <eev01q$aub$1_at_nntp.fujitsu-siemens.com>
>
> Interesting point. On the face of it, any 3VL is not a boolean algebra
> because 3 is not a power of two, and all boolean algebras have
> a power of two elements. I've long been appreciative of the fact
> that, for example, the truth table for AND is 9 cells in 3VL instead
> of four for 2VL. And the fact that while there are only 16 distinct
> binary functions in 2VL, but, uh, crap. What's that number again?
> Oh, yeah: 19683 distinct binary functions in 3VL.[1] So the complexity
> goes up a *lot.*
Never mind, it's manageable. I've done chip work with nine valued logic. Different strenghts of 0 and 1, plus undefined (ISTR different strenghts too), high-resistance and other magic stuff. We did manage to turn out quite a few working chips and so do other VHDL/Verilog developers.
Date: Thu, 21 Sep 2006 23:26:18 +0200
Message-ID: <eev01q$aub$1_at_nntp.fujitsu-siemens.com>
Marshall schrieb:
> Bob Badour wrote:
>> Can we agree that the algebra of nullable<boolean> is not boolean >> algebra and is not 2-valued logic?
>
> Interesting point. On the face of it, any 3VL is not a boolean algebra
> because 3 is not a power of two, and all boolean algebras have
> a power of two elements. I've long been appreciative of the fact
> that, for example, the truth table for AND is 9 cells in 3VL instead
> of four for 2VL. And the fact that while there are only 16 distinct
> binary functions in 2VL, but, uh, crap. What's that number again?
> Oh, yeah: 19683 distinct binary functions in 3VL.[1] So the complexity
> goes up a *lot.*
Never mind, it's manageable. I've done chip work with nine valued logic. Different strenghts of 0 and 1, plus undefined (ISTR different strenghts too), high-resistance and other magic stuff. We did manage to turn out quite a few working chips and so do other VHDL/Verilog developers.
I guess, one of the reasons the debate still rages on with no "victory" in sight is, that none of the developers have enough problems with this to be motivated for the change. Betamax was the better system, right? So was OS/2, the 68000 and so probably is a RDBMS with NULLs replaced by clever default value handling.
Lots of Greetings!
Volker
-- For email replies, please substitute the obvious.Received on Thu Sep 21 2006 - 23:26:18 CEST