Re: Schema variables - new implementation for Postgres 15

Hi

> - For sessionvariable.c (in 0002), I see that there are still all the
> comments
> and code about checking type validity based on a generation number and
> other
> heuristics. I still fail to understand why this is needed at all as the
> stored datum should remain compatible as long as we prevent the few
> incompatible DDL that are also prevented when there's a relation
> dependency.
> As an example, I try to quickly disable all that code with the following:
>
>
>
I am sending an alternative implementation based on using own int8 sequence
as protection against unwanted oid equation of different session's
variables.

This code is much shorter, and, I think better, but now, the creating
sequence in bootstrap time is dirty. Maybe instead the sequence can be used
64bite timestamp or some else - it needs a unique combination of oid, 8byte.

Regards

Pavel