Re: Normalisation
Date: Tue, 05 Jul 2005 00:17:02 GMT
Message-ID: <2kkye.1865314$Xk.720208_at_pd7tw3no>
Paul wrote:
> paul c wrote:
>
>>the protocol would be needed most obviously so that the RM would know in >>what form, for example what bit configuration, to store values or >>present them to the party that does understand those values in a way >>that doesn't undercut consistent behaviour in the future. i'd say this >>means that the RM cannot itself even decide equality. it should >>theoretically always have to inquire of some outside party whether two >>values are 'equal'. (if there is any respected definition of the RM >>that does claim to define equality, i'd like to see it.)
>
>
> I guess the problem is that a single value in a domain may have multiple
> representations. For example in a domain of rational numbers:
> 1/2 = 2/4 = 3/6 = 4/8 = ...
>
> And the relational model couldn't possibly know these are equal, because
> arithmetic is outside its area of knowledge.
>
> Unless domains had to store equivalent values in some canonical form I
> suppose? So there was always a 1-1 mapping between values and
> representations.
>
> Paul.
yes, it was a so-called 'canonical' form that i had in mind when i mentioned representation. although two different people might disagree on just which form is canonical. i suppose one could just as well call it the 'internal' form. i would think that the canonical or internal form would be specified by whatever mechanism was used for domain support. when i used the word 'representation' i was really thinking of the user or presentation or external side of things. on that side, lots of representations would make sense, such as 'localized' ones. all the domain support has to do is to make sure, for example, that what it calls equal is compatible with the psychological view that the users take.
personally, i suppose i'm lucky that English is my mother tongue because that language is one of the more dominant ones, maybe the most dominant one (i know little of Chinese, say) for people who talk about computers. still, i imagine that the RM part of a RDMBS might not have to make any concessions toward localization (i have always thought that localization should have been called globalization. then it would be more obvious how much plumbing can be involved!)
p Received on Tue Jul 05 2005 - 02:17:02 CEST