RE: Explicit Disambiguation Via RDF bNodes, more Process

>> We want a nearly-universally-shared mapping between
>> identifier strings and things they denote, but we
>
>Exactly.  We have two choices; either use URIs for this, or invent some
>other "Universal Thing Identifier" schema.  I hope we can ignore the
>third "choice", which is do some ridiculously impossible inference based
>on context, probability, and voodoo.  You need some basic level of
>confidence in your ability to identify things before getting into the
>voodoo.

I would argue that URIs are already too far gone to use for atomic thing <->
identifier mapping, and that any new "UTI" schema would be likely to go the
same way - look at the number of the ways people on this list have found to
use identifiers, references and so on. To expect an essentially linguistic,
rigid mapping to last without breaking in an environment with potentially
millions (billions?) of users seems to me to be optimistic, to say the
least. If you can give me one example in the history of humanity where such
a thing has remained constant, then I'll concede the point.

Trying to muddy the water by mixing context, probability and voodoo doesn't
help. They each have a role in different problem-solving scenarios : the
first is critical to most forms of communication, the second can be used
where only conflicting or partial information is available and the third
will cause a brief and not particularly intense pain in your left foot
shortly after reading this mail...

...however, I'm more into stuff that's do-able without magic, so I'd rather
look at the reality of the URI rather than trying to change it or invent a
magic "UTI" to replace it (apart from locally - no problem with UTIs there).
I think viewing the URI as a set and using context within inference is
probably the best pragmatic solution.

Cheers,
Danny.

Received on Friday, 26 April 2002 17:39:29 UTC