----- Original Message -----
Sent: Tuesday, September 25, 2001 9:39
PM
Subject: RE: LegalEntity Class (Was: Re:
HumanML_Write: Several responses and some reflection....)
At 10:26 AM -0500 9/25/01, Bullard, Claude L (Len) wrote:
I
suggest that if the UML models are to be pursued as a design tool, these
will have to
be the
main focus, not the schemas or RDF or any software under development.
The
reason
is of course that UML is purported to be a conceptual model from which
other
implementation mappings can be made. That's the
theory. The dilemma is that
these
can easily be considered as object-oriented designs, not data designs
as
we
discovered in Phase 0.
I think it ought to be the other way round. I prefer making the UML fit
the Schemas and RDF. As of now we have no conflicts and its just easier that
way since the datatyping is already well underway in the Schemata.
We did a
few rounds of discussion in Phase 0 on this subject, but now that
there
are
fresh members with different backgrounds, this should be revisited.
The approach
from the
Schemas perspective is to create a set of types that can be toolkits
for
creating
HumanML documents/messages. For this to work, they have to
be
usable
by any implementation of the UML conceptual layers. A schema, per
se,
is not a
class design though one can construe that by considering
abstract
types as
class definitions, but even then, these really are document
designs
or a way
to define a namespace of markup that can be added into an
existing
document
namespace. Today, the schema draft collects types from
the
semiotics field (eg, signs, symbols, etc. for describing
meaningfulness
and some
categories for environment, chronemics, proxemics, as well
as
gestural types (haptics)) and a hodge of identification types or
surface
characteristics. These are useful but there is no
implication about the
implementation or processing of these. However, without
some very
explicit
notion of how HumanML markup is to be used, I think we have
a
disconnect between the Schemas and the UML. That means,
architecturally,
we are
in the weeds and are talking around each other.
What I suggest is a very limited use of only the appropriate classes
needed to make HumanML_Write as it is currently constructed for messaging and
writing without tying ourselves up trying to make it more than it needs to be
at this stage. This would require a few new classes to fit, but not a great
deal of work.
I don't see any necessity for choosing UML over the XML and RDF
Schemata. Nothing is set in stone at this point, and if it comes down to a
choice, I think as I've said before that we should do xml and rdf and just bag
uml, except as an adjunct conducted separately, which anybody could do anyway.
I'm not saying that just to get out of a bunch more work, uml and oo, are good
for a lot of stuff, but more baggage than I think we need right now. I don't
think we are far enough along to know what kind of overarching design we need,
and you just about need to have that kind of vision to do uml well, unless you
don't mind reinventing things a few times. I mind.
However, it is good to have feedback from our newer members.
Comments
from the list members are solicited on this subject.
len
<snip>
Ciao,
Rex
--
Rex Brooks
GeoAddress: 1361-A
Addison, Berkeley, CA, 94702 USA, Earth
W3Address:
http://www.starbourne.com
Email: rexb@starbourne.com
Tel: 510-849-2309
Fax: By
Request