Kjell Rilbe wrote:

> So what's needed is a better sizecheck function that works at the
> conceptual level instead of at the implementation level.

The behavior you're seeing was not an oversight or a careless mistake.
The decision was made, deliberately and consciously to allocate string
files large enough to hold the largest probably value in the character
set and to allow as many characters to be stored in the field as fit.

Not an accident, an intentional, possibly incorrect, decision.

In the case of system tables, that behavior is essential, or all system
table objects would be restricted to 10 character names.

The problem is not that the people working on the engine are too
ignorant, incompetent, or dumb to fix what seems to you to be an obvious
bug. The problem is that the "bug" is a longstanding deliberate
behavior of the engine and changing the behavior risks breaking existing
applications. Adriano has explained a solution to the problem that
involves a new character set with new behavior. That's an upward
compatible solution that will be available in V2 - but not in the V2 alpha.

Regards,


Ann