I'm inclined to agree with Andy that the schema and orm files for each of the databases we know of should be standard and specified as part of the TCK.

There is nothing in the TCK tests that should be different in the orm files from one database to another. There are differences in the schema based on column data types, and these differences can be captured in database-specific .sql files.

The only thing we lack is broader testing on different databases. But for now, I'll be happy to require that the orm files be identical among different databases, and even happier if we can get some folks to create the xxx.sql files for databases other than derby.


On Dec 31, 2005, at 1:09 AM, Andy Jefferson wrote:

Hi Craig,

Attached please find the first draft of the TCK run rules for JDO
2.0. Please comment.

Why is the "sql" modifiable "to suit the JDO implementation" ?
Why is the "orm" modifiable "to suit the JDO implementation" ?

Surely the ORM defines the underlying schema, and so the ORM files provided 
with the TCK are totally compatible with the schema provided with the TCK. 
That is the premise we have been using with JPOX whilst developing the TCK. 
Why is it different for other implementations?
Can we have some examples of why it would be necessary ?
Are we talking about JDO implementations that don't support Apache Derby ? In 
that case would it not be better to have any other RDBMS files generated be 
fed back to the TCK project and then have them under central control? We 
can't have one JDO implementation hand crafting its own SQL files and ORM 
files and saying that it is "compatible" when the ORM they have generated may 
be incorrect with respect to the spec and the schema it should equate to. e.g 


Craig Russell

Architect, Sun Java Enterprise System

408 276-5638

P.S. A good JDO? O, Gasp!