Data Modeler: How to prevent auto surrogate key in supertype/subtypes?

Question

When forward engineering supertype/subtypes, how can I prevent the automatic creation of a surrogate key (propagated to all super/subs)? Data Modeler creates this SK only at the relational level. I have already defined a surrogate identifier in the logical; the intent is to engineer this to the relational model and propagate to each subtype.
Thanks,
Patrick

Answer 1

Thank you Philip for the explanation. I'd like to dig into this a little deeper, as I may be misinterpreting your explanation, and especially the last bit ( "...but foreign keys from subtype table to super-type table are not created" ).
My preferred super/subtype implementation consists of a supertype parent table (ideally with a [manually created] discriminator code value) and any number of subtype child tables, all referencing the parent (via FK column/RI). Note that RI is from the supertype/parent to each subtype/child table, not the other way round as Designer docs suggest (one of three different approaches: all-inclusive single-table, detail tables only, master/detail w/ RI from child to parent [arced]). The implementation I suggest ("the fourth way") allows the painless addition of a new subtype simply by creating a new child table and a new discriminator value for the parent. I've had consistent success using this approach with Designer for many years.
I've verified this successfully with a bare bones logical (3 levels), with appropriate UID on supertype/parent. All seems to transform well to relational model (i.e., all tables have inherited the parent PK, and individual attributes transformed to columns in the appropriate tables). The transform did not include a SurrogateID column.
However, the transform doesn't work quite so well in my full-blown Party model, even though I ensured that a UID was defined on the logical supertype/parent. also confirmed that Hierarchy relationships are defined between supertypes and subtypes. What else could I be doing wrong that causes the transform to create a SurrogateID column on each super/subtype table?
And on a similar note: are you suggesting that SQLDev Modeler transforms a supertype/subtype hierarchy with RI from detail tables to the parent table (i.e., backwards) when selecting "Table for each entity"? This approach would require a new FK column added to the parent table whenever a new subtype table was required.