Re: Should an application ever be allowed to change a schema?

From: Laconic2 <>
Date: Sat, 17 Apr 2004 10:15:39 -0400
Message-ID: <>

> The implication it is that part of the server application may be
> busy taking care of contraints which would have to be guarded by the
> dbms if there would be other applications. Typically, developers
> *will* implement some of the more complicated contraints in
> the server application. This severely limits the chance of the
> data ever being shared. 'Import' and 'export' facilities
> become crucial to the success of the application.

Agreed. Let me add my two US cents.

Why does the above severely limit the chance of the data ever being shared?

I think it's because constraints that are buried deep in the code are often separated from the data thus produced. Thus, if the writers of some data adhere to a contstraint that is unkown to the readers, then some information inherent in the data has been lost to the readers.

More commonly, if the readers infer that the data adheres to a constraint, when no such constraint exists for the writers, then the information the readers derive from that constraint has the quality of "lore" about it. That is, the readers attribute inferences to the data that the data does not, in fact, convey.

In practice this situation can cause things to run very smoothly, until the day comes when some data that violates the constraint is written to the data base. Then all hell breaks loose. Received on Sat Apr 17 2004 - 16:15:39 CEST

Original text of this message