Op dinsdag 7 mei 2013 11:54:47 UTC+2 schreef nvitac..._at_gmail.com het volgende:
"Normalization theory based on functional dependencies says nothing about the inter-relational constraints of the decomposition—and it can't, because normal forms are defined wrt a single relational schema."
Well, that is not entirely true, is it ?
I'm willing to bet that many a course in database normalization will effectively mention something along the lines of "and you have to introduce a foreign key constraint between the decomposed tables". Date does it too in "Introduction to Database Systems", 8ed. !
Why is that ????????
And why is it that only one of the possible two inclusion dependencies are typically ever mentioned/considered ??????? Why is it that "spurious tuples" must indeed be prevented, through the IND, from appearing in R(abcd), but that "spurious tuples" must _not_ be prevented from appearing in S(abe) or in T(bce) ???????
Because "usually", it is exactly what is desired anyway ???????? I'm inclined to agree, pragmatically, but I cannot accept that a "true scientist" would content himself with that answer. A "true scientist" would set out to seek, formally and precisely, when _exactly_ "usually" is indeed the case, and when exactly it is not.
