Re: ADR's Normalization question

From: paul c <toledobythesea_at_oohay.ac>
Date: Sat, 14 Nov 2009 08:36:36 GMT
Message-ID: <o_tLm.52824$PH1.5089_at_edtnps82>



Mr. Scott wrote:
>> ...
>> If I'm not mistaken Armstrong's axioms can be applied to show that
>>
>>   { CITY, CLASS } -> { STATUS },
>>
>> so the CTS relation isn't needed if the CLS relation is replaced by
>>
>>   CLS { CITY, CLASS, STATUS }
>>       KEY { CITY, CLASS }
>>
>> Have I got this right?  If so, I think ADR's suggested rule does apply to 
>> Date's example.
>>

>
> Sorry, you've got it wrong. {CITY,CLASS,STATUS} satisfies the multi-valued
> dependency,
>
> STATUS ->-:> CITY | CLASS
>
> which is not implied by the key
>
> {CITY,CLASS}
>
> {CITY,CLASS,STATUS} is therefore not in 4NF let alone 5NF.
> The projections, {CITY,STATUS} and {CLASS,STAUTS}, are in 5NF, though.
>
> While AB -> C can be inferred from A -> C and B -> C, that is,
> a table that satisfies A -> C and B -> C also satisfies AB -> C, but
> A -> C and B -> C cannot be inferred from AB -> C.
> ...

Thanks, I shouldn't have dropped those two dependencies and I take back the strawman crack.

But I don't get why { CITY, CLASS, STATUS} isn't 4NF, a similar MVD could be claimed for any relation that has at least one key attribute and one non-key attribute. Received on Sat Nov 14 2009 - 02:36:36 CST

Original text of this message