Re: what matters is what's said, not what's meant

> I fail to see how requiring uses to be logically consistent with privileged
> prior uses is significantly better than use implying consent (read import
> here if you like) to privileged prior uses.  In my opinion it is generally
> the case that if you don't want to disagree with privileged prior use then
> you want to consent to this previous prior use.

I think there's a big middle ground which switches sides between these
two options.   To take the running-code example I posted earlier,
requiring consistency just means 
     :Coconut a bio:Cat, bio:Dog
would be flagged as web-inconsistent (assuming bio says Cat and Dog
are disjoint), but it does not mean anything for querying systems.  It
doesn't mean you have to give --closure=po or anything.  It makes this
whole following links thing stay optional in querying, and only be
mandated (more or less) in consistency/error checking.

> To pick on my favourite example, if I want to discuss a particular invoice
> and I don't disagree with the statements about that invoice made by the
> creator of that invoice, say for example to claim that the invoice is
> invalid in some way, then I almost always want to consent to these
> statements.

I would approach that problem with some kind of scare-quotes.  I would
probably do the same in natural language: "That thing you sent me
isn't even an invoice!  The 'invoice' was just a napkin with some
chicken scratches!"  If I leave out the single-quote marks, the
overall statement starts to look inconsistent to my eye.  

(I spent a while last night trying to show how to do this in RDF, how
log:uri isn't the right kind of quoting, but....  it's not good enough
to send.  It is tricky, but I think it's doable.)

       -- sandro

Received on Thursday, 9 October 2003 16:06:27 UTC