Re: A plea for peace. was: RE: DAML+OIL (March 2001) released: a correction

(long rant; triggered by rather than targetted at Aaron's msg)

On Thu, 12 Apr 2001, Aaron Swartz wrote:

> pat hayes <phayes@ai.uwf.edu> wrote:
>
> >>> All objects? There are objects which have no name in any human
> >>> language.

> >> That doesn't prevent you from giving them one.

> > That doesnt prevent it, but there might well be circumstances which
> > do prevent you from giving them one. Obviously we can, and often do,
> > refer to things that we cannot possibly give a name to (since we
> > cannot identify them), eg as when we say that a beach contains
> > hundreds of billions of grains of sand, each consisting largely of
> > silica. It is (literally) impossible to give a name to every grain.
>
> Is it really impossible? Hmm, how about we say that I define:
>
> http://aaronsw.com/games/2001-04-12/sand_n
>
> where n is the number of the grain of sand on the shore of Lake Michigan,
> USA, numbered depth first, from top to bottom and left to write currently
> existing at the precise moment in time that I click send on this message.
> Now I've given them all names. Sure, it may be extremely, extremely
> difficult (I'm not sure that it is actually impossible) to go from one of
> these names to the actual grain of sand, but that does not mean that I have
> not given them all a name.

This is getting far too silly! The problem is that URIs are a social
fiction that've worked well enough for people, but when you look closely
we don't have a clear model of naming. So it doesn't do any good to
proclaim alice-in-wonderlandishly that some URI names some set of real
world entities. URIs were a neat hack that made web version 1.0 work
extraordinarily well, but sometimes people on www-rdf-interest seem to
treat them as magic. They're not, and the asignation of URI names to
things is just as philosophically problematic as any other kind (ie. very,
when you dig into the detail).

The nice thing about URIs is they're easy to parcel out, there's plenty
of them, and they help us do joins in databases. But if I say that
http://danbri.org/123541 names the 14th thought that Pat Hayes had today,
where does that get us. Have I assigned meaning to a URI? What about
http://danbri.org/345234234 which I now proclaim as referring to "the
number you first thought of". Silly. URIs are a convenient fiction; I
don't find much value in debating how many URIs one can assign to grains
of sand.

The vague notion of "giving a URI" to something isn't something we're
going to find easy to formalise. URIs, we say, are a special kind of name,
used on the Web. As such they share characteristics with other ways of
naming. Philosophers have killed a *lot* of trees chasing this particular
holy grail. There are no uncontroversial theories of reference: the
connection of names to the world is a deeply social, ill-defined business.
Rummage through the Philosophy of Language, or of Science, literature.
[quick google search: eg see http://ruccs.rutgers.edu/~joncohen/reference_syllabus.html ]
A great many smart people have written a great many books and papers
presenting and defending theories of how names and the world are
connected. *And we have no way to judge which (if any) of them got it
right*. Or whether the notion of a 'correct' theory of reference is a
healthy goal in the first place [Stephen Stich has written nicely on
this; don't have ref to hand]. All we know is that it is hard. And that
meanwhile, the Web still mostly works.

I fail to see how abiding by a particular syntax for names (RFC 2396)
exempts us from this the complexities that have always surrounded any
attempts to reason precisely about naming and reference. It's just plain
hard. The more we can persuade people to use URIs in the same way, the
better our Web information systems will get. But please let's not
over-sell the merits of URIs.

All of which is a grumpy way of saying that there is no merit whatsoever
in arguing about whether *all* objects are 'in principle' (which
principle?) URI nameable, since we have (and I claim will *never* have) a
clean account of naming. There are no facts of the matter about how many
names an object has, whether those names are URIs or not. But the Web
still works...

Dan


--
mailto:danbri@w3.org  (W3C/ILRT)
http://www.w3.org/People/DanBri/

Received on Thursday, 12 April 2001 14:37:06 UTC