David,
I agree with Arvola on his remark. Also,
I think that implementations can benefit
from using Role values, and I have not
heard of any MSH implementations breaking
from either having Role element present
or not having it present. (which means
that they seem to be interpreting the behavior
as a zero or one cardinality feature).
I would think it would be very bad
if a MSH said "Unsupported" because
a header had a Role element present or because
it was absent. That is why many of us
want the word "OPTIONAL" out of there.
All this debate is really reinforcing
the point of those people--like Marty,
Chris, and Doug--that we must be much
more careful in the spec about the
use of the RFC2119 key words. If _we_
are having trouble because of them,
it can't be good from the implementer's
point of view.
Dale Moberg