Re: AW: AW: Distributed querying on the semantic web

From: "Leo Sauermann" <leo@gnowsis.com>
Subject: AW: AW: Distributed querying on the semantic web
Date: Wed, 21 Apr 2004 10:42:06 +0200

> > > this implies that you are domain owner of www.whitehouse.gov if you 
> > > make up this fictous resource.
> > 
> > Why?  What requirement is there on me that I cannot make up 
> > URI resources from domains that I do not own?  What 
> > expectation is there that only I make up URI resources from 
> > domains that I do own?  This seems to go right back to the 
> > days of pre-HTML/HTTP hypertext, where there was a 
> > requirement that I had to only use valid hypertext links.
> 
> You may annotate existing uris and publish triples about them, no
> problem. You do it all the time when using RDF schemas like DC or FOAF.
> But when you make up something, use your own namespace. If you don't,
> the owner of the namespace may get very angry (Namespace
> grabbing/stealing) or may decide to use the uri for something completely
> different and then your triples are bogus.

Yes, there are risks in coining URI references in domains that I do not
own.  

I don't see these as different in kind, and I don't even see them as much
different in extent, than the risks in using URI references from domains
that I do not own.  For example, assuming that
http://www.whitehouse.gov#GeorgeBush is coined by the owner of
http://www.whitehouse.gov, even just my use of it in what I might consider
to be a sympathetic context may trigger anger in the owner of the
namespace.  The owner may also decide that the URI reference should have a
different meaning (e.g., switching between two different presidents) in
which case my use may become incompatible with the intended meaning.

> f.e. you must not use w3.org uris for your schemas if you are no w3
> developer. 

Must not?  Why not?  Yes, there can be negative consequences, which in some
cases make it infeasible to do this, but I don't know of an actual
Web-based prohibition against this.

> This is not pre-html/http age, this is one of the three
> pillars of the web: uri, http and html. 

Yes, agreed, and there are parts of the WWW where the proper functioning of
the web depends on particular documents being available at a particular
location.  Violating some of these requirements can mean that your
documents will not be accepted as valid.

> The namespace use of domains is
> common in XML namespaces. Thats how we ensure that no one fumbles in my
> namespace definitions, the internic is my legal and business insurance
> that no one steals my namespace.
> The social practice of not domain stealing is here today and not 5 years
> ago. I remember some discussions about domain stealing, people where not
> happy about it, also in this list.

Yes, but I'm not arguing that a domain owner should be prevented from doing
whatever he or she wants to do in his or her domain.  I'm just arguing that
I should be able to use vocabulary in a domain without being required to
buy into whatever a domain owner does or does not do.

> > > F.E. I have stated on 
> > http://www.gnowsis.com/wiki/GnowSis/LeoSThesis 
> > > to use this 
> > uri to identify the thesis, although not the pdf can be 
> > > found there. This may be good social practice, to tell people what 
> > > uris to use for identifying concepts. State on your homepage which 
> > > resources are identified.
> > 
> > Again, why?   There may be perfectly good reasons to do otherwise.
> 
> Because if you want a global semantic web using consistent URIs you have
> to tell people which URIs to use and use them yourself (dogfood) and
> then hope that they spread so we all talk abou the same things. Isn't
> this the idea of Semantic WEb and identifying things/concepts with URIs?

Yes, it is true that communication works better when URI references are
widely used.  Further, it is a good idea to provide information
for these URI references at standard locations.  You have ``done good''.

However, I do not see that it thus follows that I should only do good this
way.  You may have neglected to mention (for whatever reason) that there is
an early version of your thesis that you submitted to another institution.
I may want to use the URI reference
http://www.gnowsis.com/wiki/GnowSis/LeoSThesis#earlyversion to identify
this early version, and to make comments about it.  

In some sense I am being a better Semantic Web citizen by doing this than
if I made the same comments about this early version using a URI reference
that I controlled.  At the very least, I am indicating that the resouce
that I am identifying is in some sense controlled by you, even though I
probably did things this way because I realize that you would strongly
disagree about some of things I would be saying about the early version of
your thesis.

> > > I would recommend that you are only allowed to make up URLs using 
> > > domains you own!
> > 
> > I recommend otherwise.  There are many good reasons to use 
> > URI references that are not know to have been used before and 
> > that are not in domains that I own.  
> > 
> > For example, consider a situation where the WordNet 
> > information has been put in the Semantic Web, say at 
> http://realwordnet.org (wordnet.org having already been taken by a
> religious broadcasting organization), and URI references like
> http://realwordnet.org/#sailor are used in the RDF document that
> contains the WordNet information.  I would like to say something about a
> word that is not in WordNet (yet).  Why should I not be able to use a
> URI reference like http://realwordnet.org/#Bushism?
> 
> This is a good example. I agree with you that it is possible to do so
> and good practice. It allows you to do quick fixes for missing words.
> Normally, institutions like wordnet or a wiki have a social process when
> and how new words are entered. Perhaps you have to fill in some form,
> answer an email, do the limbo, whatever. You may be trying to avoid
> bureaucracy, which is ok. Perhaps this aspect is an advantage over
> existing practice and can improve our way of communicating.
> But a wiki/wordnet like URI space is an easy example. Their URIS are (by
> agreement of the publishers and social practice in all wikis) always
> identifiers for words or concepts.
> 
> On most homepages I have URLs like http://cm.bell-labs.com/cm/cs/
> You won't be happy when I say
> <http://cm.bell-labs.com/cm/cs/> <foafCorp:member>
> <http://cm.bell-labs-com/cm/cs/PeterFPatelSchneider> 
> and then publish this triple everywhere. Especially if all other people
> do so, with slightly different URIs. They may mix up with other uris you
> want to use. In a global semantic web, with the possiblitly that your
> triples end up in my aggregator, I would hesitate to use other people's
> domains.

Yes there are risks in doing what I propose.  However, again, I don't view
these risks as significantly greater than using someone else's URI
references at all.

> W3C won't be happy when I invent a new RDF vocabulary, use their
> namespace and change the meaning. Right now is a discussion in
> RDF-Calendar where Dan Connoly wanted to use HIS OWN URI in a new way
> and people where not pleased:
> 
> (iCal discussion on www-rdf-calendar)
> At 5:04 PM -0500 04.4.14, Dan Connolly wrote:
> >Summary: my discomfort with our timezone design motivated
> >me to implement a new one. Who likes it? Who dislikes it?
> Masahide Kanzaki replied:
> >Please change namespace uri if you would introduce such a 
> > significant change that makes most >existing RDFical files invalid.
> 
> so you aren't even allowed to change the meaning of your own URIs :-)))
> don't try to change the meaning of other people's uris?!

Well, this just illustrates the risks of using someone else's URI
references, in particular without some way of easily retaining a divergent
interpretation of these URI references.  In some sense, however, I don't
have much sympathy with Dan Connolly here, as he is not proposing to change
the indented meaning of *his* URI references, but is instead proposing to
change the intended meaning some W3C URI references.

> As with social constructivism: everybody has his own view of the world,
> if you express yours you can use public terms (=agreed uris owned by
> other people) or if you are not happy with them invent your own terms
> (=own uris from own domain). 

Well I want to use URIs owned by other people in ways that might not be
compatible with the meaning that their owners intend, including using URIs
owned by other people that have not been used by their owners.  What is so
wrong with this that it should be forbidden?

> greetings
> Leo Sauermann

Peter F. Patel-Schneider
Bell Labs Research

Received on Wednesday, 21 April 2004 06:15:36 UTC