I also mean that not all character states can be grounded in reality.

I agree

with Peter that they _should_ but there is a lot of legacy data out

there and

we don;t have the collation rules that were used by the authors to

come up

with

the generalities. So we need to support the word "ovate" as a leaf

shape

descriptor <<leaf_shape>ovate<</leaf_shape> or we will not be able to

support

legacy data. Even if we do not really know what they meant by

"ovate."

Of course we have to be able to <underline>handle</underline> such data, it will simply be a case of caveat emptor.

P.