Re: a URI is a name (tel uri scheme and VCARD RDF)

At 16:43 24.11.2002 -0800, Paul Prescod wrote:
>Alexander Jerusalem wrote:
>>  Now when we talk about someone's homepage for example, I don't think
>>it is appropriate to use the URL that allows us to physically retrieve
>>the homepage as the unique name of the homepage. The two roles should be 
>>strictly separated. The reason is that I could have made a lot of
>>statements about the homepage that use this URI as a unique name, that
>>is as their subject. The physical address, however, can change.
>
>
>If the address of the thing changes, then according to web infrastructure, 
>it is basically a new thing that happens to have the same content as the 
>old thing.
>
>  * http://www.w3.org/Provider/Style/URI.html
>
>The word "physical" is used in a funny way in computer science 
>discussions. There is no such thing as a physical URI. They are all 
>interpreted by server software so that there is no reason in principle to 
>_ever_ change a URI except for limitations of implementation technology or 
>effort.

Of course there is no such thing as a physical URI but a URI can be used in 
some cases to physically retrieve data. And in those cases it contains 
technical information as to how to retrieve the data (in the form of a 
protocol specifier). I think a very good case can be made for unique names 
to be kept free of any information that carries meaning.


>Anyhow, why not use RDF to make the assertion that the new page is 
>daml:equivalentTo the old one (or be more precise and say something like 
>the new URI supercedes the old one).

That'd be a kind of versioning scheme for identities. But although this is 
certainly a valid approach I think it's more sensible to keep identity 
separate from any other properties a thing may have such as how to retrieve 
it. In fact the latter shouldn't even be seen as a property of the thing 
itself much less as a part of it's identity.


>>So even for internet resources, I think we should have one URI to name
>>it and another one to point at it, even if the two happen to have the
>>same sequence of characters at any one point in time.
>
>
>Discussed here:
>
>http://www.w3.org/DesignIssues/Axioms.html#Properties


This paper shows quite well the contradictory notions of what a URI should 
be. On the one hand it says URIs should not change and be opaque. On the 
other hand meaning is encoded into URIs, such as protocols and hierarchies 
and different parts are extracted and acted upon by different agents. The 
topic of what should be in an ID has been discussed at great length long 
before the web was invented (i.e. in the context of relational databases) 
and I tend to be on the side of meaningless unique IDs cleanly separated 
from information about how to get the stuff the ID refers to.

Received on Monday, 25 November 2002 12:23:30 UTC