Q1: What is the purpose of normalization?

A1: To reduce data redundancy.

Q2: Can we remove all redundancy?

A2: No we cannot; we need to maintain minimum redundancy.

Q3: What is the minimum redundancy"

A3: The minimum redundancy to related tables - primary key - foreign key relationship, or "referential integrity".

Q4: So the only redundancy we can have is the PK-FK columns?

A4: Yes. So, if a field/attribute is not PK or FK, it may not appear in two tables. In other words, all feilds must appear in ONLY ONE table, unless you're PK-FK.

Q5: Can a table/relation not have a field that also appear in another table/relation?

A5: In that case, how your table/relation relate/join with other table/relation? So every table/relation must have a field that also appears in another table/relation.

Q6: When examining a given data table for functional dependency, how do I decide when a composit key jointly determine another field ("Full Dependency")?

A6: You can use the first field in the composit key ALONE to try to see whether it alone will determine the values of the said field; if it does, then this is in a dependency because only part of the key is sufficient to determine the other field's value - so these two fields involved are in a partial dependency. On the other hand, if one field in the composit key ALONE cannot determine the value of that other non-key field, if it needs BOTH fields in the composit key to work together to determine the values of that non-key field of interest, then this (these three fields) is a full dependency.

For e xample please see Normalization steps and demo questions, the three columns of SalesPersonNumber, ProductNumber, and Quantity.

Q7: If I identify, say, four FDs in the original 1NF relation (table), after normalization to 3NF, how many relations (tables) will I end up?

A7: The answer is obvious, so I'll leave it to yourself to make :-) If you refer to the document "Normalization rules/tips (Version 4)", you can find the answer on P.3, bullet 10.