Re: PSVI

[Rick Jelliffe <ricko@allette.com.au>]

From: Henry S. Thompson <ht@cogsci.ed.ac.uk>

>I think to claim to do so would be at best misleading -- you _can't_
>assign types without knowing that higher (in the tree) type
>assignments were correct, i.e. local validation and type-assigment go
>hand in hand.  Perhaps I've missed something.

Are you saying that if there is a local declaration for an element (e.g.
x:x/y ) but the validation for x:x fails (perhaps due a previous required
element being missing) then y _must_ use any global declaration
for y?   That does not seem to be clear to me: is that the status quo?

I thought one reason for the "no single element name with two types" rule
was to overcome that, so that schema assessment of children could continue
even in the event of failure of the parent to validate?

Surprised,
Tapei