home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.databases
- Path: sparky!uunet!news.uiowa.edu!herky.cs.uiowa.edu!bonak
- From: bonak@herky.cs.uiowa.edu (Esmail Bonakdarian)
- Subject: Normalizing 2NF -> 3NF
- Message-ID: <1992Nov15.163413.17527@news.uiowa.edu>
- Sender: news@news.uiowa.edu (News)
- Date: Sun, 15 Nov 1992 16:34:13 GMT
- Reply-To: bonak@herky.cs.uiowa.edu (Esmail Bonakdarian)
- Nntp-Posting-Host: herky.cs.uiowa.edu
- Organization: U of Iowa, Iowa City, IA
- Lines: 40
-
- I have a question about normalizing a relation from 2NF to 3NF.
- My understanding is that we try to remove any sort of transitive
- dependencies during this step.
-
- E.g., given a relation with three attributes, A, B, and C, where
- attribute A is the PK, the following dependencies exist:
-
- +----+-----+----+
- | A* | B | C |
- +----+-----+----+
-
- A -> B
- B -> C
-
- i.e, A determines B, and B in turn determines C.
-
- 3NF would split this up into two relations:
-
- +----+-----+
- | A* | B |
- +----+-----+
- and
-
- +----+----+ +----+----+
- | A* | C | OR | B* | C |
- +----+----+ +----+----+
-
- Which of these two latter relations is the "correct" one according to
- the theory? Or does it always depend on the given semantics of the data
- being modeled? It seems to me that I've seen both.
-
- Could somebody please clear this up form me? Also, any references to
- material/articles (both on/off the net) that explain the normalization
- process clearly would be much appreciated.
-
- Thanks!
-
- Esmail
- PS: If this isn't the best place to ask this question, please tell me
- where to post this instead. Thanks
-