Re: Axiom: Opacity of URIs


If a naming authority (the person or organization who assigns the  
URIs) documents and publishes a rule for deriving their URIs, then of  
course everyone is allowed to use this rule to derive URIs.

This doesn't violate the opacity axiom at all. The axiom is about the  
opposite direction. If a system finds a URI somewhere, then it should  
not make any assumptions about what the URI refers to. All power over  
what a URI refers to should be on the URI *producer* side, not on the  
consumer side. If, for example, your client system is designed to  
treat all URIs ending in .jpg as images, then your system violates  
the opacity axiom, because it grabs power away from the URI producer.  
The producer gets to decide what is an image, not the consumer.

The problem disappears if the client acts with the consent of the  
producer. (E.g. the producer has stated, "all URIs at the  
domain that end in .jpg are images. This policy will be in effect for  
the next five years.")

The TAG finding Danny pointed to discusses these issues at length.


On 4 Jan 2007, at 19:42, Ronald P. Reck wrote:

> I have a question about URIs. I think I understand the axioms here:
> My question concerns the Axiom: Opacity of URIs
> When I work with words/strings I want to say something about them.
> Now, imagine the English word "pain".
> rdf:about=""
> Its real helpful to be able to derive its URI so that every time I  
> come
> upon the same string I dont incur the "pain" associated with asking  
> some
> system the question: do you know the URI for the string "pain"?
> Now, imagine I am processing French and I also have the string "pain".
> Well, I believe its intrinsically a different word so I think it  
> makes sense to change my URI structure to
> rdf:about=""
> Attractive but wrong?
> Wrong only when I look for French words with "//fre" ?
> I run into the same situation when I want to manage definitions for
> words from multiple communities of interest. The word "frequency" has
> slightly different meanings in the domain of mathematics, physics or
> signal processing. Again, it is attractive to have derivable URI's.
> Can someone please comment or point me at relevant disscussion in  
> this area? -thanks.

Received on Thursday, 4 January 2007 20:30:46 UTC