Hi Lester,

>
> > I have just finished importing a 19 Gbytes database to Firebird

1.5

> > (over 70 milions records). Yes it's a real world application.

And

> > yes, all the queries runs in under one second.
>
> From What ?

There are 4 tables. Each of them contains some diferent combination
of information. For example: One have phone, address, SSN, etc. The
other has some car information, like License plate, owner, address.

The user fill up the information, and the client aplication
will "try to find" as much information as it can using those 4
tables.
This is taking less then a second ;-)

>
> > Another feature that would be great: Linking Database.
>
> If the files are on the same machine then it could be feasible,

but then

> people will expect it to work across machines ;)
>

I know that this feature across machines would really be very
usefull. But we should at least have some linking. So if it's
possibe to have same server, diferent databases link, it would be a
major step. I guess it would be easier to develop, and a lot of
people (including me) would be able to use and test it.

> > Although I am trying to avoid using it, select count(*) from

table,

> > should really be improved :-(
>
> That IS the correct way to do it. The count is only valid for the
> transaction that accesses it, and only you know how you are

counting

> records. As well as a total count ( or just SUM sub totals ) it is

very

> useful to keep sub-total counts for different groups of records,

and

> only you know how to do that :)

The fact is that when I am modeling my tables, I do a lot of
thinking to improve Agregations performance, this way I can always
use SUM, AVG , COUNT(*). The problem is that the user needs are
followed by some nasting "excel" formulas that involves counting a
lot of unrelated datasets, and that evalutation ( counting the
records for that dataset) take too much time :-(
Examples are Forecast, Budgets, Sales...


Clément