W3C home > Mailing lists > Public > www-rdf-dspace@w3.org > December 2003

RE: ungetable http URIs

From: Seaborne, Andy <Andy_Seaborne@hplb.hpl.hp.com>
Date: Fri, 5 Dec 2003 14:18:54 -0000
Message-ID: <E864E95CB35C1C46B72FEA0626A2E808712DE8@0-mail-br1.hpl.hp.com>
To: "'Kevin Smathers'" <kevin.smathers@hp.com>, Nick Matsakis <matsakis@MIT.EDU>
Cc: Stefano Mazzocchi <stefano@apache.org>, SIMILE public list <www-rdf-dspace@w3.org>



-------- Original Message --------
> From: Kevin Smathers <mailto:kevin.smathers@hp.com>
> Date: 4 December 2003 17:43
> 
> I think that in your argument for didacticism you ignore the real
> problem of when resources should be gettable, and when they should point
> to real-world objects.  An OCW course lists the contributors to that
> course; gettable or ungettable.  Your argument doesn't provide any
> advice -- it depends on what you mean by a contributor, a real person,
> or a proxy for that person.  I think we all agree already that real
> people cannot be delivered by an HTTP GET.
> 
> The problem is just pushed down a level.
> 
> I propose that the databases we have in hand always reference proxies,
> not real things.  That is, when OCW lists an author, what it really
> means is an entry in its list of authors, not the author him or
> herself.

This is saying that the entry in the database is a reference to a controlled
term for the author - you can GET on that URL and get the controlled term
information.  That works; it is natural language that is loose in talking
about a person and the concept (controlled term) about a person.  We all
just say "person" for both.

>  Likewise works of art, &c. all reference not the actual work
> but what we know of the work.  Assigning URN's to real-world objects is
> the task of an authoritative index such as the Getty ULAN, or the OCLC
> name authority; 

Here I am not so sure.  Anyone can give a name to a thing.  Using one from
an authority (in issuing controlled terms in thrie vocabulary) is mor
euseful as it will be shared more widely but anyone can create a name for a
thing.

> unless we commit Simile to the work of being such an
> authoritative index (a task which there has been no whisper of), the
> best we can do is reference the entry in our records of who the person
> is, and try to link that record to the authoritative references.  In
> other words, ALL of the references we work with should be gettable URL's

This is certainly workable.  Whatever we decide, we are not just making
techncial choices.  There are several versions of the web architecture: each
is consistent, workable and usable but they do behave differently.

	Andy

> 
> Cheers,
> -kls
> 
> 
> 
> Nick Matsakis wrote:
> 
> > On Mon, 1 Dec 2003, Stefano Mazzocchi wrote:
> > 
> > 
> > 
> > > > I think things that don't meet those requirements should get
> > > > ungetable URIs. 
> > > > 
> > > > 
> > 
> > 
> > 
> > > What kind of URIs wouldn't meet those requirements, IMO? [not
> > > caustic, just curious] 
> > > 
> > > 
> > 
> > I think its generally pretty clear whether things can be expressed as
> > bits or not.  JPEG image? bits. Oil painting on canvas? not bits. 
> > There are some digital artifacts whose bits are constantly changing
> > or even never the same, such as websites, but those are still bits
> > (in fact, HTTP 
> > can provide metadata for those bits that lets you know whether they
> > are static or dynamic).  I don't have any objection to things that
> > can be expressed as bits being given gettable URIs, even if they are
> > not available at those URIs. 
> > 
> > The other fringe case is internal nodes in RDF data structures.  I'm
> > not sure what the current consensus on this is, but on systems that
> > implement b-nodes they can be blank, and not have any URI.  If, for
> > some reason, they do need URIs assigned, then I think ungettable ones
> > would be appropriate. 
> > 
> > Nick
Received on Friday, 5 December 2003 09:24:06 EST

This archive was generated by hypermail pre-2.1.9 : Friday, 5 December 2003 09:24:09 EST