We don't currently say what the effect is of setting type="xs:untyped" on instructions such as xsl:element.
I propose that we define it to have the same effect as setting validation="strip".
Affects both 2.0 and 3.0.
The WG accepted the proposed interpretation.
We also need to consider type="xs:untypedAtomic". The decision here is that we should validate as for type="xs:string" (that is, check there are no element children) and then mark the node as xs:untypedAtomic.
Changes applied. Test cases added: validation-0107, -0108, -0109.