James Clark wrote:
> I suspect you can always transform positive literals out:
>
> <element><choice><name>a</name><name>b</name></choice> p </element>
>
> =>
>
> <choice>
> <element><name>a</name>p</element>
> <element><name>b</name>p</element>
> </choice>
Right. But if we have ten such nameclasses
<group>
<attribute><choice><name>a1</name><name>b1</name></choice></element>
<attribute><choice><name>a2</name><name>b2</name></choice></element>
...
<attribute><choice><name>a10</name><name>b10</name></choice></element>
</group>
we have
<group>
<choice>
<attribute><name>a1</name>p</attribute>
<attribute><name>b1</name>p</attribute>
</choice>
<choice>
<attribute><name>a2</name>p</attribute>
<attribute><name>b2</name>p</attribute>
</choice>
...
</group>
If GNF normalization is applied to this pattern, we have combinatory
explosion (2^10) again.
Cheers,
Makoto
