On Thu, Apr 28, 2011 at 3:07 PM, Karsten Hilbert <Karsten.Hilbert@xxxxxxx> wrote: > I must agree with a recent poster > that what appears to identify as a natural key often really > isn't or else becomes not so later on. It's vastly easier to > then deal with that by re-defining constraints without > having to touch primary keys. I have run into this. Often what happens is that the data model directly affects the assumptions coders make. In a system I worked on, the username was unique / primary key. Code was written that assumed username would always be unique across the app. Then the app was deployed to multiple non-associated groups, where user names from two different organizations might be the same and they couldn't be changed. Talk about some ugly code changes needed to be made to fix all the bugs. ugh. An initial assumption that a serial value would be unique, but usernames might not be would have resulted in a much cleaner design for that system. Or even that the PK was org||username or something. -- Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
