[
Lists Home |
Date Index |
Thread Index
]
So you imposed a code system on top of a code
system and expected everyone to notice that
different rules apply depending on the code
system in effect?
Namespaces are a code system (URIness for uniquifying)
built over a code system (GI names in a domain vocabulary).
Problem is, XML GIs don't qualify fully as
a code system either: no rules for linguistic
interpretation, just syntax and has-a. No
IS-A, no episodic rules, etc. These have to
come from other systems, so as long as we
know that, we can make do.
One solution is to do as the RDFers
are saying and create IS-A rules as metadata.
I don't think that solves all problems but
it entirely depends on the types and contexts
of communication. Humans use time and space
episodically, (see Schank's MOPS), so predicate
logic isn't quite enough for human semantics, but
it is one tool. Human KR needs more than
predicate logic. Yes, Paul, the SW isn't
intended to be human KR, but it is possible
that the SW isn't the last word on the subject.
Why URIs aren't "words": if they were, they
would be signs in a code system and sets of
structural and linguistic rules would be
available for building them into combinations
or sentences. In other words, one could
construct "treatments". The namespace rec
is right insofar as they are a syntax device
for disambiguating the codes expressed in
the instance (one can sort name from name
and know they are different signs), but they
do not offer a linguistic rule for intent:
why choose THIS code in this context from
all other possible choices in this context,
or why was THAT code not chosen. Analysis
of communication behavior requires analysis
of the choices made.
Semantic noise comes of:
1. Competence. Is the code being used
and interpreted correctly (in accordance
with linguistic and structural rules)?
2. What is emphasized in the treatment?
3. What attitudes are conveyed to the
receiver (eg, dominance, an example being
Tim's insistence that he can tell
whining from facts, thereby asserting a
dominance relationship)
4. What attitudes are held toward the
message (does it have value, to whom,
in what context).
<not_to_John>
Attitudes are predictors of behavior. If
you want Tim to fix it, approach him with
the right attitude (an analysis of the message
says he likes dominance, so obsequious might
work). If you don't care, don't bother.
On the other hand, a sign of planning behavior
indicates a probability of action. So the
fact that he is here and making signs means
he is willing to work the problem. Maybe even
wanting but still waiting.
</not_to_John>
len
From: John Cowan [mailto:jcowan@reutershealth.com]
Tim Bray scripsit:
> I'm on the developer's side. Looking back, the namespace of unqualified
> attributes is just a goof in the namespaces REC. I wonder if there's
> any chance of fixing it?
I doubt it.
I think the Namespace REC got this right: the point of namespaced attributes
is for the equivalent of attribute architectures, and they just aren't the
same thing as plain ol' element-local attributes.
|
