At 09:40 AM 3/28/00 +0200, you wrote:

>From: Christer Matson <christer@...>
>
>Well, people have give the computers the instructions how to accomplish the
>smartness in the first place. And then the productivity gains must be

greater

>than the cost of this smartness. So I guess one very important point is

having

>the optimizer do a more than decent job without too much overhead, which

isn't

>the case in many situations as it stands today.
>
>I think your question below is too general, of course people are needed to
>make computers smarter.
>

Of course. But if a developer does it right, once, then a 100,000
users don't have to do it every day. A good payback ratio.

What I'm trying to do is establish a cultural mindset for Interbase
development, soon to be a world wide effort, that easy to use means
intelligence in the system, not thousands of individual features.

Open Source has produced some very high quality software but often
at a cost of great obscurity. Samba, for examples, emulates what
on Windows takes three mouse clicks, but requires 140 pages of
documentation to read and understand before installation. Of
the "info" documentation reader with so many easy to remember
commands that they ran out upper case letters, lower case letters,
control-letters that they had to resort to keyboard chords.

I tried to make Interbase as simple and fast as possible -- no
tuning (other than indexes), no configuration control (undone
by Sys-V shared memory!), no overrides. Where tuning was
traditionally required, I found alternative algorithms that
we insensitive to tuning. Although memory sizes now make
the feature unnecessary, the optimizer computers the number of
page buffers required to efficiently perform a query and
warns the cache manager to find more memory.

I want Interbase developed by people who believe that smart,
small, and simple is good. I want it used by people with
a problem to solve but who don't want to be database experts.


Jim Starkey