Geoff Worboys wrote:

>
> That was the point of my comment "except that we dont seem to
> store the increment as part of the generator definition".
> If the increment was part of the generator definition then
> the "standard" would be defined. That aspect is not feasible
> with the current implementation.

It's a subject of extension. The standard allows user-defined increments
defined at the DDL level. We just need to extend the system tables to
store the extra properties. After that everybody who has USAGE privilege
(also to be implemented) can call NEXT VALUE and only an owner of the
generator can alter the current value.

The major question is whether we should still allow full-featured
GEN_ID() calls (as a legacy insecure stuff) or treat any non-1 increment
as ALTER and hence require the developer to adjust his code to the new
rules.


Dmitry