Re: Normalization, Natural Keys, Surrogate Keys

From: Tobin Harris <>
Date: Sat, 18 May 2002 15:26:33 +0100
Message-ID: <ac5o6i$mh3n1$>

> The reason I like singular is that when you do data model validation
> on the cardinality, you speak of an instance/row, not of multiple
> rows. This is why I wouldn't agree with people who say to use plural:
> [order] -----> [order_detail]
> (0,N) (1,1)
> "An ORDER may or may not have an ORDER_DETAIL"
> "An ORDER_DETAIL must have at least one and at most one ORDER"

That's exactly the same reason I use it. It's the cardinality that determines whether something is plural or not. For some reason though, Celko (ISO) uses the number of records a table contains. It a table contains one record then it would be singular, otherwise plural, or a collective noun preferably. I'm pretty sure this is how he described it.

I wander if it's better to use singular for logical modelling, then plural for physical modelling? It the mo, I use singular for both, until someone can convince me otherwise.

Tobin Received on Sat May 18 2002 - 16:26:33 CEST

Original text of this message