OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.


Help: OASIS Mailing Lists Help | MarkMail Help



   Re: [xml-dev] URIs and Names on the Web

[ Lists Home | Date Index | Thread Index ]

apologies to Joshua Allen / Jonathan Borden for my late nite posting/rant,

since I criticized ( waited a few cycles as pointed out by Simon St.
Laurent ), I should also comment ( putting on flame retardent flight
suit... ), but I can't help feeling that this is how other victims were
trapped....... I did indeed go through the vast number of emails when trying
to assimilate this topic, so I am certain I have missed main tenets of the
arguement, here is a gelogical analogy as inspired by Didier Martin's

a new rock speciman, found in the field, could possibly be named;

- using a latin name

- using the name of the person who found it

- using the name of the place where it was found

- using the name reflecting characteristics of crystalline habit, chemical
compisition, hardness test, color etc...

- a mixture of the above

To be sure, different geologists from different parts of the world will have
different names, possibly reflecting their own, possibly simultaneous,
discovery of the rock or perhaps they will simply have a different name due
to their language. Presumably their could also be a different naming schema
because this new rock 'fits in' with some previous findings or ontological
framework. Not to mention that there will be groups of geologists
considering one name more 'truer' then another. Names also can change
through time....language mutation, context, etc....

To make things even more difficult, here are some other 'real life' naming
and classification scenarios;

- 2 rocks of exactly the same composition may be named because of the
process that created them were different ( metamorphic .... )

- an artificialy created stone may be given a different name, even though it
matches the exact chemical composition of another rock, there is also a
'brand name' of certain rocks.

- I have even heard of a rock that was named after the nearest 'sound' where
it was first found; which was that of a waterfall, so it was simply named
'waterfall' though this was in the local tribal dialect ( which was spoken
by less then a few thousand people ! )

so we could state that;

- a name does not automatically imply 'meaning' about itself, but it might
( e.g. someone named John Carpenter, may have had a carpenter as an
ancestor )

- a name is not automatically an address or locator, or a universal token of
identity, but it could be, especially as defined in the terms of scope and

- the actual binding of a semantic 'handle' to an inanimate object, occurs
quite easily ......( imagine me pointing to a hat made of cheese....and I
utter 'cheesehat', we can now have discussions about cheesehats ) as
illustrated in the above naming situations.

- a name does not neccesarily need to conform to some ontological naming
system, other then possibly the rules of the underlying language

I am definately out of my depth, with respect to any kind of 'logic'
formalisms....but it does strike me that creating ontologies is partially an
effort to encode within a name some sort of further meaning about itself. An
ontological system also possibly defines a sense of place or 'where' any
element plots respective of other elements in a classification system.

This is useful if you are a student trying to remember 3000 different kinds
of e.g. rocks....but what is the analogy with machine manipulated ontologies
? Do they need hints other then those which assist in processing? Of course
we humans still need em.

A problem I see is the mixing of the transport layer with the semantic layer
? I don't see why we need to bind or create any formalisms to specifically
HTTP, other then where performance demands it. Though admittedly we need to
use existing HTTP methods to achieve what we want, so ( as proven by Thomas
Passin very recent RFC analysis this looks sensible, possibly even
foreshadowed ) I don't see why we can't have named resources via HTTP,
though we could admit that the use of them in an HTTP derived semantic web
would have to be explicitly defined, and thusly the questions and statements
we could pose to such systems ( and their related responses) would have
dependancy upon these definitions.

As illustrated by the above blah blah on geology, an admittedly small scant
sampling of human knowledge,that its almost a 100% certainty that a resource
will probably have many different name(s), identity and locations throughout
its semantic lifetime ( hehehehe, just put semantic infront of it; and
presto new word ). A world which constrains a named resource to one
universal convention, isnt a world I want to live in. Though my example
focused on the 'name' aspect, I think that there has been enough about
identity and locations previously. I think having a trully 'universal'
naming convention is something not achievable and possibly counterintuitive
to what is desired. If I was a perfectionist I would have a real issue with
the idea of using  HTTP naming scheme to address non-network-retrievable
resources, but since I'm not I can dismiss it  is not as not being terribly
relevent ( yet ).

to possibly sum it up;

'Whoever or whatever that is involved in creating statements or having a
semantic conversation over HTTP using URL's as a named resource(s), should
consistently check and agree on the convention being currently employed for
the conversation(s) about the use of named resource ( and maybe even related
scoping rules for identity and location ), so instead of an implicit builtin
HTTP bounded relationship that a URL IS the 'name'of a resource. We could
have a little bit of negotiation that works out the contract that states the
ground rules ( with of course a default contract that reflects the most
common situation of URL as a name to a resource )'

in real life, what does this mean, I could imagine a well formed xml
document gets exchanged | referred | annotated | updated between peers,
clients or servers, in the process of creating a statement | question |
answer on the HTTP enabled semantic web. The result is that a statement |
question | answer would refer to this contract and thusly have a dependency.

The vocabulary of the xml document would presumably have to explicitly
annotate conventions and scope to be used in the statement | question |
answer. This should definately exclude any notion of mapping.

and who cares about the performance implications.....seems to fix itself
every few years any how !

cheers, jim fuller


News | XML in Industry | Calendar | XML Registry
Marketplace | Resources | MyXML.org | Sponsors | Privacy Statement

Copyright 2001 XML.org. This site is hosted by OASIS