At 10:30 PM 10/28/01 -0000, jwcanada@... wrote:

>I am not sure if this is right place to post. IB/Firebird is one of
>the best in term of easy to manage, stability and standard feature
>support.
>
>One thing is that always forces me to think about moving to other
>database such as MS SQL or ORACLE is the query speed in IB/firbird,
>especially with sub query.
>
>As our application needs to support all major industry database (MS
>SQL, ORACLE, db2 etc) in the same code base, it is difficult to use
>IB-Specific "NON-STANDARD" PLAN feature to speed it.
>
>Any plan from atchitect point of view to addres optimization issue. I
>believe that it shoudl be the top priority after firebird1 out before
>adding new features.
>

The whole concept of plans is a crock -- a wart on a wart,
an abomination, a crime against software, an feable excuse
of a feature to mask a bug in the optimizer.

How would anyone feel about a C compiler that required the
programmer to tell it how to optimize loops? Or an automobile
that required a trip profile before it could adjust the
ignition timing?

Software can and should optimize itself. It knows more about
its innards that application programmers. It can estimate
costs, cardialities, and alternatives. It doesn't need
a human to tell it what to do. Firebird was designed to
make the optimizer cheap, simple, and effective. And it
was until Borland broke it.

Fix the optimizer. Then dump the entire concept of plans
forever.


Jim Starkey