James Clark wrote:
> <choice>
> <attribute name="foo"/>
> <attribute name="bar"/>
> </choice>
>
> is the same as
>
> <attribute>
> <choice>
> <name>foo</name>
> <name>bar</name>
> </choice>
> </attribute>
>
> I think if you allow one, you should allow both. I don't see any problem
> with:
I can certainly relax the restriction. I wrote:
> The subordinates of each <oneOrMore> pattern has either a single
> <attribute> pattern or a <attribute>-free pattern. Otherwise,
> the pattern does not have a GNF.
A <oneOrMore> pattern satisfies one of the two conditions.
Condition 1: neither <element> nor <ref> occur as a (possibly non-immediate)
subordinate.
Condition 2: <attribute> does not occur as a (possibly non-immediate)
subordinate.
Note that I am assuming your normal form.
Cheers,
Makoto
