>I real world applications, the 3NF sucks. I don't even want to prove this statement as there are plenty of profesional articles (at least in dutch) which say this.
>Yes it 'll give you some information: if the count in the stock table is not equal to the count in the articles table it indicates that a bug or crash occurred. from this you may can resolve the problem.
>
>As for the performance matter a single table query is faster than a multiple table query for this matter.
>
>My intention was to tell how the optimizer works (in general) and was telling a theory. The samples were brought in by other people. To test my theory i did some testing with real-world data which i cannot send (for the same reason) through the internet.
>
>Note that RECCOUNT is also a form of redundancy (like the stock field in the articles table). I rarely want to know the exact number of records.
>
>If i have a process which scan a table a processes record by record i'll use a stratgy like this:
>
>nRec=RECCOUNT()
>SCAN FOR whatever
> ......
> nProgress=RECCOUNT/RECNO() * 100
>ENDSCAN
>
>note that this will only work is no index order is set.
>
>I still don't see much use of what you're trying to tell me.
>
>In fact mostly they won't need it.
>
>Like i said to markus,
>
>The deleted feature is in fact only a field in the table. If we don't add an index to all fields in the table because some indexes on those fields might not
>be of any (or only a very small) benefit to performance, why should you add the deleted tag ?
>
>I tend to use as less indexes as possible, which improves replacements and appending data. Having more indexes could lead to more index corruptions (you cannot corrupt indexes which don't exist), and as we all know, some kind of corruption is very hard to detect.
>
>Only in cases where performance is a really important issue, i'll take the proper steps to enhance performance by adding an index (might be an index on deleted()).
>
>I'm saying here that we must choose our indexes carefully. This is far more effective than just adding just an index because the optimizer may use it sometime.
>
>Walter,
3NF=Third normal form, as in normalization of a table, right? It strikes me what you say about this. 3NF is minimal requirement, IMHO. I've been following both -index on deleted()- discussions, and it really seems that you don't like the established practices. "The establishment", let's say. All along you say that this is a theory of yours, but you talk as if it was 100% proven. But then again you also say 100% proof is hard to get. The fact that you've got a hypothesis about something doesn't make it a rule for everyone else, and that is not the message that has come through (at least for me) when you defend your arguments.

Regards,