>>>>> "Dimtre" == Dimtre Novatchev <dnovatchev@xxxxxxxxx> writes:

    Dimtre> Even the last E2 in a path expression should evaluate
    Dimtre> either to a sequence of nodes or a sequence of atomic
    Dimtre> values, but not a mixture of the two... I find this
    Dimtre> unreasonably restrictive

Is it really so restrictive in practise?

Given that E1 must be a node sequence, then even if the restriction
were not there, you would still only be gaining functions of type

[a] -> [a|b]

where a is node() and b is atomic().

Is that so useful?
-- 
Colin Paul Adams
Preston Lancashire