Mike Matrigali wrote:
Daniel John Debrunner wrote:
...
My second concern is that many places get characters and the change
must ensure they get the correct datatype, apart from potentially
being a lot of work, the chance of missing some or picking the wrong
character types seems high.
What is the plan for system catalogs? I just realized that if system
catalog char's are going to be non-collation based then I guess 10.3
would really already be supporting 2 different collations per database.
I think this is why I expected new type id's. I didn't expect the type
id's to represent a specific collation, just that it had a non-default
collation. Thus system catalogs would always have default collation.
So it seems like we need to choose:
1) system catalogs share locale based collation with rest of db
2) add new type ids as previously proposed
3) support per conglomerate collation
note that we need to support proper recovery of system catalogs so we
can't solve the problem though some magic runtime manipulation of system
catalog queries.
