Re: test cases for fromUnionQuery, please

On Thu, Jun 02, 2005 at 08:03:58AM -0400, Kendall Clark wrote:
> It also makes for lumpy (i.e., inconsistent) semantics, since FROM
> <service-provider-uri> doesn't mean the same thing as FROM
> <any-other-graph-uri>; or, if it does mean the same thing, it's an
> unnecessary constraint on the service provider, which is now required to
> publish the triples it wants every query to be executed against (in addition
> to any other ones the requester specifies) at some URI.
> 
> > It seems reasonble to me that the default behaviour should be to trust
> > whatever the service trusts, overridable by specifying given graphs. After
> > all, the client is passing the request to a specific service to answer. 
> 
> Hmm, I think talk of "trust" is misleading. But, otherwise, I agree. That
> is, using my earlier example, I don't trust myself, I simply believe that p,
> and I want p to be one of the claims that is in the background when I answer
> queries. Other, *pure* query answering services may not have any background
> beliefs, which is fine.

Agreed, as soon as I typed ZZ I realised I shouldn't have used the T word.
Sorry. I was using it as a macro for all-the-stuff-to-do-with-provenance.

- Steve

Received on Thursday, 2 June 2005 12:47:30 UTC