Dmitriy L. Kulakov wrote:
I think Jackrabbit is too young for such issues.
ORM doesn't work at all in a version that a have checked out two weeks ago.
It is interesting to have a look at db schemas in orm contribution.
There we can see a rather common approch of mapping xml-structured data to
strict static set of tables.
You can also look at Edgar's purely JDBC implementation here :
http://issues.apache.org/jira/browse/JCR-91
Granted it is a bit young, and because Jackrabbit is still under
development, the "optional" PMs have been lagging behind. The idea was
to get the Jackrabbit implementation to version 1.0 as soon as possible,
and the contrib would lag a bit behind because they are not regularly
maintained with the rest of the project. Maybe adding some documentation
to explain this would be a good idea.
I rather like the idea of having the database set up in a
similar way to
how the JSR-170 pec suggests (with one table per property
type) for type
2 repositories. It's more "normal" for systems people and is also
available to systems outside JSR-170 (non-Java - MS ADO for example).
So general implementation should do DDL for definition of new nodetypes and
mixins.
I think it is rather difficult. But interesting.
I think that in the early days we should mostly use JSR-170 as a
high-level API than JDBC, that offers us powerful searching,
hierarchical types definitions, referencing, etc, than as a general
mapping tool.
Regards,
Serge Huber.
