> No single feature would make me happier than a multilevel name space.
> The question is how to implement one without rewriting the database from
> scratch.
>
> Here is a modest proposal, more or less off the top of my header:
>
> 1. We increase the size of RDB$RELATION_NAME from 31 to 256 bytes
> (and make it a varchar, this time, please).
> 2. Schemas are undeclared, just used (this is topic for debate, folks)

Please make it declarable. This would give people a bit more direction
in how to use and how to present schemas in development GUIs.

> 3. A qualified table name is represented in the system tables as
> <schema>.<tablename> as in "myschema.mytable"

I would rather have a separate column for the schema an object belongs
to. Hey, why not use RDB$OWNER_NAME for that? Or will there
be a difference between owner and schema an object belongs to?

> 4. When fetching tables, we parse the table name, look for the dot,
> and consider the prefix the schema name, the suffix the table name

See above.

> 5. In BLR we recognize and parse table names containing dots as
> qualified names
>
> Almost nothing changes except the max length of table names and the name
> matching rules in the BLR and SQL compilers.
>
> It seems too simple. What am I missing?

Martijn Tonies
Database Workbench - tool for InterBase, Firebird, MySQL, Oracle & MS SQL
Server
Upscene Productions
http://www.upscene.com
Database development questions? Check the forum!
http://www.databasedevelopmentforum.com