Re: Normalization and Derived Information

"Tony" <andrewst_at_onetel.com> wrote in message news:ed8a00fa.0410130105.2ffc8e2a_at_posting.google.com...
> "Laconic2" <laconic2_at_comcast.net> wrote in message
news:<UIednTgiwvcvd_bcRVn-jw_at_comcast.com>...
> > A body of data that stores every fact once, and only once, is an
auditor's
> > nightmare. Because if, through error or fraud, somebody manages to
store
> > an incorrect fact in the database, and that incorrect fact damages one
> > person and benefits another, there isn't any way (without stepping
outside
> > the database) to determine that the fact is incorrect.
>
> Yes, I see what you mean. But surely the redundancy is only really
> more trustworthy if it comes from different sources? Merely recording
> the same fact (or derived fact) multiple times as part of the one
> transaction initiated by the one user shouldn't improve our confidence
> in it, should it? Or is the point that a malicious hacker modifying
> the data through the "back door" is likely to fail to change all
> copies, and hence expose the hack?

You're getting deeper into accounting than I really want to go. I'm not sure that the terms "hacker" and "embezzler" cover exactly the same ground. If you're interested in pursuing it further, you might want to pick up a book on accounting and look at the section under "Accounting System Controls". Or you might want to look at some of the disasters that have occurred when accounting systems were built by IT people who were insufficiently aware of the need for "controls".

Getting back into IT for a minute. When a packet traverses a network, its generally suffixed with some kind of cyclic redundancy check (CRC). Is that redundant? You bet! If anybody started yelling at me about CRC's being denormalized, I would think that he was suffering from what the French call "idee fixe".

In fact, DEC Rdb used to have a checksum at the end of every database "page" (a fixed length block in a file). If anybody "patched" some data without correcting the checksum, the system could detect the corruption. I don't know about Oracle or the others. Received on Wed Oct 13 2004 - 13:33:21 CEST