Re: A Normalization Question

Neo wrote:

> I realize the following are extraordinary examples, however a general
> data model can't (application above it can) have prejudices as to what
> data/changes it will accept. Suppose, the world is taken oven by
> islamic terrorist. As part of their spoils, they want every word in
> every computer to be spelled backwards, thus 'brown' needs to be
> update to 'nworb'. Or suppose, the French, take over and want every
> string to end in a silent t. In the above tuple, updating one and not
> the others, corrupts the db. Below is approximately how XDb1
> normalizes the three strings and updating it from 'brown' to 'nworb'
> or appending the symbol t does not corrupt the db.

Oh yes, it does, in exactly the same way the relational model suffers from this, uh, "problem." Suppose a database contains two "browns" and one "green." Given your requirement, It is just as much a "corruption" to change both browns and miss the green as it is to change one brown but not the other. Therefore, tokenizing each individual string does nothing to solve your "problem" here.

Larry Coon
University of California Received on Thu Jul 08 2004 - 23:48:49 CEST