> I was also going to mention that we almost always have to de-normalize
> SQL databases to perform optimally and quite frequently have to build
> special tables with completely redundant information just to do things
> quickly. While everyone seems to think SQL is the be-all end-all, my
> practical experience over 10 years dealing with them has proven
> otherwise. If you keep things simple, everything works fine. As soon as
> real world complexity intrudes, things get nasty quickly. Why do you
> think there is a multi-billion dollar industry in things like
> wharehouse's and datamarts?
yep. I think we speak the same language. Denormalization in order to increase
performance lurks evrywhere and is very tempting. BUT: it complictaes the damn thing,
makes it error prone, M$ like (a nicely polished surface with a lot of shit under the
hood). I do it, but I don't like it a bit. I have the feeling that �there must be some
intelligent solution, like what Boyer-Moore did for full text search or B+ trees did
for indexing. Such a common problem in modern GUI software, and no patent solution yet
... Drives me mad every day.
And yep again, SQL RDMS servers are not ideal, but still the only thing available here
and now that does the job for our purposes without investing another few man years in
writing our own database kernel.
Horst
