[
Lists Home |
Date Index |
Thread Index
]
If you don't mind, John, I'd like to detach cleanly from
the capitalism thread at this juncture and focus
on applications of RNG and RNG AF. As Rick Jeliffe
notes wisely, we are pursuing understanding which
applications will benefit from using RNG. We
are told that RNG is best for documents and
XSD for databases. Because I often generate documents
from databases, that may or may not be an exclusionary
answer.
Because you are familiar with this effort and RNG, you can
probably explain most clearly why RNG is a good
candidate for the base. My lack of clarity makes
me think I would need XML Schema for the base and
could use RNG for the secondary schemas. XSD provides
abstract types and it would seem these are a useful
means to organize the element types that can then
be used in a renaming system, or can be carried by
"class" attributes.
Marketing: if we establish undeniable value, we
can find market. It helps to know where to look for value.
Companies like Microsoft are in the business of making
computer systems ubiquitous and know now that ease is
very important to that. So regardless of who sponsors
RNG, if it has value, I believe it will be implemented
in commercial software. Ease of uptake by medium skilled
users is very valuable. We know from experience
that if we set out to create a language by which humans
communicate with other humans and computers, and our
efforts result in a language that is hard to learn
and hard to apply, however well the computer likes
it, we goofed. Schema designs based on toolkits
must be very easy to use because of the innate
difficulties of working with novel or unfamiliar
abstractions in the domains themselves. A complex
tool on top of that is deadly.
So that leaves me in a quandary and I need help
with this. (Your help is very valuable and I
know your time is, so many thanks.) But maybe these
problems are exemplary of where and why XML developers
need RELAX NG. A case made by example is illuminating.
Finding the right example is a challenge.
In HumanML, we've managed to a certain degree to isolate
out the major categories of characteristics of people,
environment, and channels that affect the success or
failure of human communication acts. That semiotics,
hermaneutics, etc. were our sources is no great surprise.
So we have a pretty good set of top domains for glueing
together the application languages of different groups
that want something like HumanML. But a glue language
is exactly what we will have to begin with. It has
to be simple to apply because it is a toolkit.
At the same time, I need to think about an implementation
in which the HumanML data is just a database, possibly
relational, and used to feed rendering applications or
to create messages that the Human agents can act with.
len
-----Original Message-----
From: John Cowan [mailto:jcowan@reutershealth.com]
I would have to see details, but I think using RNG for the
base schema would make a lot of sense.
|
