Re: lcsh.info RDFa SKOS and content negotiation - use of RDF-style # IDs in RDFa?

On Oct 6, 2008, at 12:13 PM, Pat Hayes wrote:

> On Oct 6, 2008, at 6:12 AM, Steven Pemberton wrote:
>
>>> How does not giving the concept a name solve Dan's problem?
>>
>> The aim of a name is to uniquely identify something. As far as I am  
>> concerned, as long as I have a method of uniquely identifying  
>> something, I'm happy.
>
> What would make me very happy would be for someone (perhaps  
> yourself?) to tell me what is meant by 'identify'. Take a concrete  
> example: suppose I have a name, told to me by someone, and to keep  
> things general lets suppose that this name is not a URI.  Now take  
> two cases. In the first, this name is simply a character string. In  
> the second, it also identifies something, in your sense of  
> 'identifies'. What is different about these two cases? What can I  
> do, using the name, in the second case that I cannot do in the  
> first? And how does anything I can do with the name relate to  
> whatever it is that the name identifies?
>
> Thanks for any help you can give.

I'm completely with you on this one. Let me attempt to agree.

One says that X identifies Y as being Z by virtue of some property P  
of Y being I.  I, which could be a string or anything else, plays the  
role of the "identifier" in such a scenario - it is Y's identifier (in  
this scenario), it identifies Y.  ("X identifies Y" and "I identifies  
Y" are two different senses of the word "identify"; this is confusing.)

For example, a railyard employee (X) identifies a particular railroad  
car (Y) as being the one (Z) from Acme Coal Co. that was filled in St.  
Louis, by virtue of examining its bar code P(Y) = I and seeing that I  
encodes the number 87753207 (which is written only on that particular  
car; and X knows this). The bar code identifies the car.

If you don't (erm) identify X and P in such a scenario, you end up  
with meaningless statements. Does "zwmx" identify something? That  
question means nothing until one has answered: To whom, as what, and  
in what way?

It may be that some marking P(Y) = I does not identify Y as being Z -  
there may be some other Z' for which P(Z') = I, so you would be able  
to tell whether Y was Z or Z'. So identifierness is not a syntactic  
property of anything; it depends on knowing that such a Z' cannot exist.

Since I first heard you making the point about "identifies" abuse, I  
have been replacing "identifier" with "name," and so far it has always  
been an improvement. Well, occasionally one wants to say that the  
identifier is "87753207" and the name is "car number 87753207"; but I  
restrict my use of "identifier" to markings on or parts of the thing  
whose identification is enabled by such markings or parts.

Jonathan

Received on Monday, 6 October 2008 19:03:19 UTC