Greetings,
* Petr Jelinek (petr.jelinek@2ndquadrant.com) wrote:
> Certain "market leader" database behaves this way as well. I just hope
> we won't go as far as them and also create users for schemas (so that
> the analogy of user=schema would be complete and working both ways).
> Because that's one of the main reasons their users depend on packages so
> much, there is no other way to create a namespace without having to deal
> with another user which needs to be secured.
I agree that we do *not* want to force role creation on schema creation.
> One thing we could do to limit impact of any of this is having
> DEFAULT_SCHEMA option for roles which would then be the first one in the
> search_path (it could default to the role name), that way making public
> schema work again for everybody would be just about tweaking the roles a
> bit which can be easily scripted.
I don't entirely get what you're suggesting here considering we already
have $user, and it is the first in the search_path..?
> TBH I would personally prefer if we got rid of search_path as GUC
> completely because it makes certain aspects of DDL logical replication
> and connection pooling much more complex, but that does not seem to be a
> realistic change.
No, I don't think we're going to get rid of it.
> > opportunity to do so. I do think it would be too weird to create the schema
> > in one database only. Creating it on demand might work. What would be the
> > procedure, if any, for database owners who want to deny object creation in
> > their databases?
>
> Well, REVOKE CREATE ON DATABASE already exists.
That really isn't the same.. In this approach, regular roles are *not*
given the CREATE right on the database, the system would just create the
schema for them on login automatically if the role attribute says to do
so.
Thanks!
Stephen
