Lists Home |
Date Index |
Murali Mani wrote:
> Thanks, I read HT's mail also. The main thing I was trying to see was:
> how to define keyrefs to "relative keys" (scope is not root element); same
> scope is perfectly fine..
> if you define keyrefs on different scopes, then you run into problems.
> Important point worth double-checking and for all of us to note, I felt..
> note; we can have key on an "ancestor" scope, and a keyref on a
> "descendant scope".. i guess that is not allowed right now.. in most such
> cases, we can move the keyref to the ancestor scope.. right??
In the data model you laid out, moving the key and keyref to the
ancestor (root) scope really don't buy you anything (even after you add
the necessary state code to car).
Key/keyref seem well-suited to describing elements that can be
"flattened" into relational tuples, not so good for true hierarchies.
I'm glad to hear HT and company are working the problem. Chained keyrefs
sound like a useful and usable addition. (At least, they'll be usable
after Priscilla explains them in the second edition of Definitive XML