>> - <value> and <data> are not "nullable", ie they don't match the empty
>> sequence
>
> This is a welcome news, because otherwise it is very hard to correctly
> compute nullability in an efficient way (because a datatype can accept ""
> or reject it depending on the context.)
Yes. At the moment, I am cheating by assuming that any context-dependent
datatype cannot match the empty string.
>
> But I am still unable to confirm that the mentioned spec change will
> affect the implementation exactly in the described way.
If you can indicate where you are unsure, maybe I can explain some more.
James
