To make my code work with XSLT implementations that don't support
selection of nodes by their xml:id attributes, I replaced occurences
of id($some-id) with //*[@xml:id=$some-id], which works but for
obvious reasons doesn't make me happy.  In the source documents, there
can be a potentially large number of elements that are inter-connected
by means of their IDs.  It's a bit like in a GedML encoded family
tree[1].

Are there any recommended alternatives to the discouraged //* solution?

Thanks!
Thomas W.

[1] http://xml.coverpages.org/GedML-Sample-kennedy-xml.txt