A supply farm can have a transportation document. If present it can be one of two types: internal or external. Both documents share some common data, but have different specialized fields.
I though of modeling this in a OO-ish fashion like this:
alt text http://www.arsmaior.com/tmp/mod1.png
In the document table, one of the two doc_*_id is null, the other is the foreign key with the corresponding table.
That is opposed to the other schema where the common data is redundant:
alt text http://www.arsmaior.com/tmp/mod2.png
I’m trying to discover pros&cons of both approaches.
How do I SELECT to know all the internal docs in both cases? We have a sort of mutually exclusive foreign keys, the JOINs are not so trivial.
Is the first approach completely junky?
Classical ER modeling doesn’t include foreign keys, and the gist of your question revolves around how the foreign keys are going to work. I think that what you are really doing is relational modeling, even though you are using ER diagrams.
In terms of relational modeling, there is a third way to model inheritance. That is to use the same ID for the specialized tables as is used for the generalized table. Then the ID field of the doc_internal table is both the primary key for the doc_internal table and also a foreign key referencing the supply_farm table. Ditto for the doc_external table.
The ID field in the supply_farm table is both the primary key of the supply_farm table and also a foreign key that references either the doc_internal or the doc_external table, depending. The joins magically get the right data together.
It takes a little programming to set this up, but it’s well worth it.
For more details I suggest you google “generalization specialization relational modeling”. There are some excellent articles on this subject out there on the web.