Re: My action item on Moby Dec, issue 14, etc

On Fri, Sep 20, 2002 at 04:11:51PM +0100, Miles Sabin wrote:
> Jonathan Borden wrote,
> > A URIref is a _label_ for a _node_. Each URIref uniquely and
> > unambiguously labels the node. Not all nodes are required to have
> > labels. Nodes may have properties and these properties may relate one
> > node to another.
> This might well be a perfect description of an idealization of the web, 
> in the same way as a regualar grammar might be thought to be an 
> idealization of a natural language. But neither bear very much 
> resemblances to either the web or natural language as they actually 
> are.
> The actually-existing web isn't simply a collection of abstract nodes 
> and processes: it's a network of extrinsically semantically laden 
> entities created and maintained by people. As such it's liable to all 
> the ambiguities, contradictions and general messiness that characterize 
> natural language and other social phenomena.
> You can pretend otherwise, but if you do you'll end up with a very 
> precise description of completely the wrong thing.

I think what you're missing with that statement is that different aspects
of the architecture occur at different 'layers'. Let's take a different but
similar example: DNS. The DNS is defined in RFC 1034 and RFC 1035. Using
those and possibly a few updates you can write a DNS client and a server
and use them just fine with the public Internet. But what you end up
missing is all of the stuff that people use DNS for and how they use it:
trademark law, dispute resolution policies, registries, registrars, cc-tlds
vs gtlds, etc. Is that a problem with the DNS system? No because everyone
knows at what 'layer' each happens. The 'architecture' of the DNS is
pretty well established and comfortable with where ambiguity exists and
where it doesn't exist. At the protocol layer, labels match exactly. When
domain-names get used in browser they might be spell checked, searched,
or otherwise mangled.

The 'actually-existing web' is very much a collection of abstract nodes
and processes _at that layer_. Its when that layer gets used inside
processes and system that humans use is when the issues of semantics
come into play. Don't try and flatten the layers into each other. Layering
is _useful_.


Michael Mealling	|      Vote Libertarian!       | urn:pin:1      |                              |
!! The Trailblazer spacecraft is going to the Moon! And for just $2500 a gram !!
!! you can send something along with it! Business cards, momentos, cremains,  !!|| anything! See for details!                     !!

Received on Friday, 20 September 2002 11:32:04 UTC