Re: Putting context in RDF serialization

Seth,

Maybe I misunderstood your point.  I certainly agree that it is good to 
build higher-level constructs to ease the management of lower level 
entities.  My point is that I don't think the lower level constructs need 
to be changed to accommodate this.

Sure, I will buy a "bag of nails" rather than a collection of crystaline 
groupings of Fe atoms, (and various impurities).  But the basic atoms are 
still the same basic atoms.  Some of these basic atoms have been used to 
create containment structures (the bag) to facilitate handling.  But the 
fundamental atomic properties are not changed.

Thus, I say that RDF is sufficient for a useful range of purposes, and 
contains much of what is needed to create higher level structures.  Let's 
try and build on top of RDF rather than reinventing it.  (A notable 
exception is full first order logic;  I happen to favour the approach of 
defining this separately.)

As for what it is that you pass between systems, there is a different (and 
interesting) debate.  I think that basic RDF is a useful lowest common 
denominator, but, as you say, applications may7 find better ways.  This is 
not a debate that I regard as critically important at this time (though it 
may become so as RDF is more widely used).

[In writing this, I am reminded of a particular RFC: 
ftp://ftp.isi.edu/in-notes/rfc1437.txt]

#g
--


At 03:32 PM 1/1/01 -0800, Seth Russell wrote:
>I think you guys missed my point here.   Let me try some analogies:  when you
>go into the hardware store to buy nails, do you find each individual nail
>labeled with its size?  When a chemist studies a substance, does he name and
>handle each molecule individually?  No, these things are not practical. Rather
>we just label  containers.   The methods you guys are suggesting (sure 
>they can
>be made to work) all require that each individual statement is tagged as a
>separate entity and labeled with its context.  I would like to humbly submit,
>that people (and programmers) are just not going to do that.  Rather we need a
>method of allowing packages of statements to float around, with the packages
>labeled with their context, rather than each individual statement.
>
>So you say:
>
> > If the context can be expressed in the RDF graph-syntax (which I is the
> > approach I have tried to follow) then I see no need to create extensions to
> > the RDF serialization.
>
>Your methods seems to refer to a model of the internal structure of data 
>inside
>of an implementation.  As such they work fine.  But is it really practical for
>diverse systems to pass this internal information around externally amongst
>each other like that?   Perhaps two systems that are already pretty much in
>sync, might want to transfer a batch of statements, numbering each statement
>and then also transfer the exact context of each of the statements as well.
>But in the open ended environment  of the Semantic Web, I think we will want a
>way to put context labels on large chunks of information.
>
>Seth Russell

------------
Graham Klyne
(GK@ACM.ORG)

Received on Tuesday, 2 January 2001 12:19:55 UTC