At 09:56 AM 6/28/02 -0400, Alexandre Kozlov wrote:

>
>It looks like not perfect approach. I guess it would be better to escape
>this
>transformation for indices. C++ OOP should help it easier to design and
>maintain.
>

I hate to be the one to break this to your, but nothing in life is
perfect.

Before you go off and spend a vast amount of time attempting to
re-invent the wheel, you might to an analysis of the problem you're
trying to eliminate.

The existing mechanism is simple, robust, and correct for all
intended operations. Index walking is almost always wrong --
it does random, deoptimized access into the database page space,
is a nightmare from a locking perspective, and is almost always
much slower than the alternative.

If somebody really cared about timestamp index precision, a better
solution is to use a non-lossy transmogrification for 64 bit ints.

(I do hope everyone understands that "transmogrification" isn't
an actual English word but a reference to a beloved but extinct
comic strip.)

Jim Starkey