W3C home > Mailing lists > Public > semantic-web@w3.org > January 2008

Re: HTTP URIs for real world objects

From: Peter Ansell <ansell.peter@gmail.com>
Date: Thu, 17 Jan 2008 18:08:06 +1000
Message-ID: <a1be7e0e0801170008r2592fa8es1e165fbad496d0a9@mail.gmail.com>
To: "Peter F Brown" <peter@pensive.eu>
Cc: "Bernard Vatant" <bernard.vatant@mondeca.com>, "Danny Ayers" <danny.ayers@gmail.com>, "Reto Bachmann-Gmür" <reto@gmuer.ch>, "Leo Sauermann" <leo.sauermann@dfki.de>, public-sweo-ig@w3.org, semantic-web@w3.org, "Paul Roe" <p.roe@qut.edu.au>, "James Michael Hogan" <j.hogan@qut.edu.au>

Peter,

See comments inline.

On 17/01/2008, Peter F Brown <peter@pensive.eu> wrote:
> I don't think the issue is as simple as that. You are right about the OASIS spec (which, if memory serves correctly, didn't make it through to approval as a standard and sits a little uncomfortably as a "recommendation" that doesn't have the authority of its namesake in W3C) but that underlined precisely one problem for both camps: what authority can make assertions about the identify of proxies for real-world subjects? In practice, it doesn't matter as the recommended (but by no means mandatory) guideline for establishing "ownership" of any claim is the namespace.

Authorities own namespaces. That works for me... Real world subjects
can be commented on by anyone in a democratic semantic web. Trust
doesn't need to relate to freedom of speech. No-one can force you to
use a specific statement.

> The published subject identifier (PSId) is merely a mechanism for making an assertion that "in this neck of the woods, we define this thing this way" but with one important and significant provision: any conformant processor must assume that any two topics, whatever their names or other 'properties', refer to the same real world subject if they carry the same identifier. RDF has no such - pragmatic - mechanism. The reliance on the use of the law of identity doesn't help as it assumes that for any two things, you can only assert that they are identical if the set of their properties are equivalent, an assertion that is difficult to sustain if one does not know the entire set of properties (and, in many real-world situations, the complete set of such properties is possibly unknowable).

It depends on whether you use a logic which requires strict observance
to its rules, or one which can cope with omisions and incorrect
elements. Mathematical proofs of the completability of computing
algorithms don't need to be performed before the program can function.

If you stop people talking about things without authority you just get
ignored. If you stop people defining properties of things in a
democratic web you will be ignored. If you stop people from using your
vocabularies without permission they will define their own and ignore
you. Basically you can't stop people from making progress by telling
them they have to talk about things using an identifier that you
decided on. Until the web focuses on using "sameAs" or something more
descriptive than "seeAlso" for relationships between creations by
different users, RDF will be stuck in an age where properties won't
mean anything because they can't change or extend them in their own
publications.

The world doesn't work by figuring out exactly what everything means
and then keeping definitions intact after that. Imagining the semantic
web to agree with a contract where different identities have to be
bit-perfect in order to recognise anything about them seems to be
missing reality.

> It comes down to an issue of governance, who asserts what and with what authority, and bland assertions that the world is going to populated by happy RDF triples will not make that issue go away.

It may be surprising how far a world of triples based on RDFising
proxies could go. DNS authorities are the least democratic system we
have on the web. Why not use them?

> The subject of this thread is about URIs for real world "objects". RDF's fatal flaw in this respect is that it makes no distinction between use and mention of a subject (think Alice Through the Looking Glass and "the name of the song" - Lewis Carroll knew what he has talking about). I would still maintain that PSIds answer that requirement unequivocally, however poorly it might be considered that the documentation supports that view.

I think any assertions with the ID/URI as the subject of a statement
should be regarded as defining a property specifically about the thing
that is being referred to. If you want to do serious proofs which
require things to stay the same then you really should control
everything in your laboratory for the duration of your experiment. Our
knowledge of real world objects is free to change whenever scientific
discoveries, or any other assertion for that matter, is made. Does
this make the semantic web human?

Not sure what you mean by "use" and "mention". I would have said
"definition" and "use" in the sentence I was thinking about. PSId's
don't offer a realistic solution for everyday users in the ways they
have been implemented so far. Telling someone they have to sign a
contract on the web has never worked, and neither will asking them to
sign a contract to get into the semantic web.

If semantics were not just about academic subjects and random
relationships between researchers then we could get an idea about
whether things actually would work. Albeit reasoning in an open
environment is bound to fail, but simple operations like following
chains of relationships will work, especially if you recognise
equality at a lesser than bit-equality level so people can reform and
manufacture statements themselves.

Peter Ansell

-----Original Message-----
From: Peter Ansell [mailto:ansell.peter@gmail.com]
Sent: 17 January 2008 02:53
To: Bernard Vatant
Cc: Peter F Brown; Danny Ayers; Reto Bachmann-Gmür; Leo Sauermann;
public-sweo-ig@w3.org; semantic-web@w3.org
Subject: Re: HTTP URIs for real world objects

And to think we haven't gone anywhere since 2003, except for away from
Topic Maps it seems. It was particularly noticeable that OASIS didn't
specify how to publish specifications, only that you should if you
want to use their mechanism. The difference between the figures in
Sections 2.3 and 2.4.4 would not be recognised by a pragmatist looking
at the document as it is without any discussion of publication
mechanisms.

Before standards boards realise it there is going to be a de-facto
standards with single simple RDF http resources backed up by OWL
documents to define the vast majority of the Semantic Web.

Peter

On 17/01/2008, Bernard Vatant <bernard.vatant@mondeca.com> wrote:
>
> Thanks Peter for the reminder
>
> There is also a separate OASIS recommendation on Published Subjects,
> defining "published subject identifiers" and "published subject indicators".
> http://www.oasis-open.org/committees/download.php/2897/pubsubj-pt1-1.01-cs.pdf
>
> Bernard
>
> Peter F Brown a écrit :
> > There are HTTP URIs for real world objects: they are called "subject identifiers" and form part of the ISO 13250 Topic Maps standard, where a 'Topic' is a proxy for any real world subject, and a Topic can take 0 to a URIs as subject identifiers...
> > Its sometimes really hard keeping things simple, but when There is something There that works , please use it and don't fall victim of the 'Not Invented here' syndrome. Regards, Peter
> >
Received on Thursday, 17 January 2008 08:08:16 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 21:45:20 GMT