Re: On, and on, and on...

At 11:41 PM 2000-06-08 +0100, Graham Klyne wrote:
>At 12:16 PM 6/8/00 -0500, you wrote:
>>But I'm not sure I understand the position of many other
>>participants in this list; I continue to see
>>misunderstandings of the essential specs, such as the
>>distinction between resources and entities[1], and
>>I find it worthwhile to (try to) clear these up[2] and
>>understand the arguments better.
>>
>>
>>[1] Graham Klyne's message of Thu, 08 Jun 2000 09:05:56 +0100
>>http://lists.w3.org/Archives/Public/xml-uri/2000Jun/0362.html
>
>That was a slip on my part, which unfortunately seems to have obscured the 
>point I wished to pursue.
>
>I'll re-phrase my comment:
>
>Ignoring, for now, issues of relative and context-dependent URIs.  If a
>namespace is a resource, and a namespace name is a URI:  what resource is
>identified by that URI?  Logically, it is the namespace (which may be an
>abstract, non-retrievable entity).  But if one chooses a namespace name
>that can also be used (directly) to retrieve some schema bound to the
>namespace, then the resource identified by the URI ipso facto is the
>resource represented by the schema document thus retrieved.

It would be proper ["good software engineering," not "Rec compliant"] to
refer to the schema and not the namespace when it is not proper to process
the instant document in light of the namespace without regard for the
schema, presuming that the schema tells you everything you need to know to
process the names that come under its purview.  [I'm not saying you can't
parse it, but the document is dependent on the schema and the intended or
"highest and best" application of the document depends on schema
conformance.]  

If one cites the schema as one locally declares the namespace, a
connotation that would seem natural is that names bearing this mark locally
are to be interpreted in accordance with the strictures in the cited schema.

The framers of the Namespaces Recommendation did not want to commit the
processors that they contemplated building to a responsibility for a
potentially limitless cycle of reference, recovery, and interpretation of
schemata.

This is behind the history where this very un-specification-like "it is not
a goal" language went in-band, and out of band stronger language to the
effect that "the namespace has no semantics but its identity."  This is all
in an ontological space of out-of-band assumptions where lots of divergence
is possible because the in-band provisions of the Recommendation[s] just
don't say.

The "abstract all interdocument references via the URI class" policy, which
says documents referring to resources are not to narrow their expectations
other than that the resource cited has a URI, bears a lot of the blame for
why this has not been more effectively spelled out in the specification
literature.

Normative references, acknowledging a higher authority, need much more
control or safety provisions than to hyperlink href's.  In another parallel
universe Nick Ragouzis and I developed the idea "the browser 'back'
function is_a 'undo.'"  There is no 'undo' for a normative reference to a
schema.  So the forward path has to provide for your safety.

The kind of control one needs on this relationship link is radically
different from what is needed for hyperlinks from which there is an easy
recovery.

I am an advocate that there should not be a unique language or MIME type
that is acceptable at the end of the "normative reference to [partial]
language definition" link.  On the other hand, the idea that there should
be no class restraint other than that it have a URI is [searching for
polite language].


>
>It seems to me that this resource represented by the schema document must, 
>in general, be different than the namespace resource;  in RDF terms, I can 
>make statements about it (who created it, etc.) that are not statements 
>about the namespace.
>
>Where now the 1:1 correspondence between URIs and resources?
>

Yes, this is the point that both David Carlisle and Rich Jelliffe have
pointed out:  there is a need to be able to distinguish the identity of
these things at times.  What computer scientists can have a hard time
digesting is that there is also a need to undistinguish them at times.  The
solution has use for names or identifiers which are precise and those which
are less precise.

The point that the namespace purists are missing is that we were never
talking about abstract namespaces in the first place; what we are talking
about is a key to a sub-language derivation: some text which is XML and in
addition there is this additional knowledge about the names of its element
types and attributes.  We are discussing _namespaces in XML [markup names]_.

>...
>
>Another view, contemplating the quote you shot back at me:
>
>       "Resource
>          A resource can be anything that has identity.  Familiar
>          examples include an electronic document, an image, a service
>          (e.g., "today's weather report for Los Angeles"), and a
>          collection of other resources.  Not all resources are network
>          "retrievable"; e.g., human beings, corporations, and bound
>          books in a library can also be considered resources.
>          The resource is the conceptual mapping to an entity or set of
>          entities, not necessarily the entity which corresponds to that
>          mapping at any particular instance in time.  Thus, a resource
>          can remain constant even when its content---the entities to
>          which it currently corresponds---changes over time, provided
>          that the conceptual mapping is not changed in the process."
>
>         -- http://www.ietf.org/rfc/rfc2396.txt
>
>This admits an electronic document as a resource, of which a schema 
>document is an example.  Then what one gets doing a GET on its URI is an 
>entity containing a representation of that schema document.
>
>...
>
>Or, treating the resource as the conceptual mapping from URI to entity, I 
>perceive two such:
>Schema as resource:
>  - the mapping to a document created to describe some properties of some
data
>Namespace as resource:
>  - the mapping to a definition of properties associated with the namespace 
>(of which a schema may be a part).
>
>...
>
>What does it mean, in terms of the conceptual mapping involved, to say that 
>a namespace is a resource?

>

The namespace is an abstract object that has a proper method which applies
to tuples {name, namespace} where 'name' is an entity type name or
attribute name in XML and 'namespace' is the abstract object we are
defining, and this method returns values comparable to the "is x an element
of S" query from the classical Boolean algebra of sets.

We have been having an ontological problem where purists say that a
language which satisfies all these clauses is "not a namespace" because it
has proper knowledge beyond that laid out in the specification of the
'namespace' class.

This is where I have been trying to build an ontological bridge between the
two camps by casting the namespace as a 'view' of property of the language.
 The language can be abstracted to its proper namespace, if [like XSLT] it
has one.  This is viewing, or projection, of the intension of the language.

Languages and graphs, the WWW, are subclasses of not-necessarily-finite
continuing structures.  Many of these continuing structures have a
finite-form definition in the form of a grammar or other schema.

Namespaces belong to the space of ancestors of these structure classes:
less completely specified relatives of network classes where the network
classes include classical graphs and trees of computer science, as well as
markup languages which have a structure in this tree or graph form plus
connotations expressible as constraints or relation applications across the
nodes identified in that structure.  

From what I hear we have adequate substantiation for both the desire [XSLT]
to have a namespace which is the captive of some schema, not [desirable] to
be used without following the strictures of that schema; and in other cases
the desire to have namespaces [HTML] which are used commonly under a
variety of schemata. 

This information could be recorded in a schema-level-schema as part of
describing the operational requirements surrounding the use of namespaces
in XML.

An interesting issue that has emerged recently is that XML applications
need name partitions that are of limited scope as well as XML-global spaces
of names.  At least I find this interesting because I find the notion of a
"global name" as a primitive or axiom impossible.  The virtue of URIs is
that they give you a big enough locale so that identifiers qualified for
use in that scope enjoy a domain of definition big enough to reach and be
shared among all the other participants in the Web.  If beings on Mars are
communicating with beings on Antares, we haven't de-conflicted URIs with
regard to what they are saying.  But we have built a local scope big enough
for our practical [data] communication capabilities.

Al


>#g
>
>------------
>Graham Klyne
>(GK@ACM.ORG)
>

Received on Friday, 9 June 2000 09:57:42 UTC