- From: Pat Hayes <phayes@ihmc.us>
- Date: Thu, 15 Apr 2004 01:53:31 -0500
- To: Jeremy Carroll <jjc@hplb.hpl.hp.com>
- Cc: Patrick Stickler <patrick.stickler@nokia.com>, Chris Bizer <chris@bizer.de>, www-archive@w3.org
- Message-Id: <p06001f7cbca3a76bcdd6@[10.0.100.76]>
>Pat > >part 1: >I think we made progress ... > swp:assertedBy . swp:authority > is true when the authority asserts the graph, >part 2: > and from the point of view of the authority, this includes the >truth of the graph itself. Yes, but it is hard to express 'point of view' formally. > >(note we had that when the authority provided the graph, but that is >not needed, if Pat asserts `Jeremy asserts "foo bar"', then for me, >`Jeremy asserts "foo bar"' is true if: >a) I asserted "foo bar" >and >b) "foo bar" > >even though you (or anyone else) might be providing the graph. Yes, your objections were good and I had to think it all through again. Let me summarize my own current understanding informally first (as much for my own benefit as for y'alls) . The way I had it, a warrant graph asserting G entailed G, which was wrong. The warrant graph just says that the authority asserts G, but G doesn't follow from that alone (since I can believe that without believing that G is true) . I had been looking at it all from the asserter's point of view, not the reader's point of view. (What might be the case, plausibly, is that if you trust the authority then you should believe what the authority asserts: that is, G is entailed by the conjunction of an assertion that the authority is trustworthy and a warrant of an assertion of G by the authority. But this inherently involves trust. Without some trusting going on, you can never formally derive an asserted graph from an assertion of the graph by any authority. (You might not even trust yourself in some extreme circumstances.) ) We assume three notions as primitive: that of an 'authority' which is some kind of legal or social entity which can be referred to by a URI (so is the 'person' in the domain of discourse) , and a relationship between such authorities and graphs which we will call being the authority of, or authorizing, the graph, and the idea of a warrant. Exactly what these mean is not specified formally, but intuitively the authority is the entity which accepts responsibility for standing behind what the graph 'says'. We suggest two techniques for detecting the presence of the authority of a graph and of an authorization relationship: by being published on the website belonging to the authority, or by being signed with a signature of the authority. Authorities can also stand in some kind of propositional-attitude relationship to a graph, such as asserting it, denying it or quoting it. A warrant is a thing that connects all this together: a warrant graph describes it (so is true or false depending on whether the relationship of authorizing holds or does not hold) and can also under some circumstances, acting as a performative, actually make the relationship it describes be true. To do this the authority it describes has to really be authorizing the graph in question. In this case the warrant graph actually *is* a warrant (given the presence of the authorizing relationship) and so is sufficient to ensure the truth of the warrant description considered as a piece of RDF. This allows other kinds of warrants to exist, but still gives the performative force. Being asserted/quoted/whatever by a self-describing warrant is just true by fiat: if you say you are asserting G, then you are asserting G, etc.. But the formal story ends there, and we only get connections between assertions of G and G itself by invoking a trust policy. ------- Changes to the text attached in TEX format which I am slowly learning (they will need some formatting work for the footnote and figure, but I hope they are not completely hopeless.) There are a few comments inserted in TODOs: delete after reading. I figured we had to rewrite section 5.1 and 5.2 as well as the formal stuff in 6. These are intended to be complete replacements for the text in these subsections, so if something is omitted that is deliberate. Those files are also on my website, in case the attachments get screwed. Sorry it took me so long to grok what y'all have been talking about. Pat PS. In section 4 we say "There are currently two ..." Aren't there rather more than two, in fact? -- --------------------------------------------------------------------- IHMC (850)434 8903 or (650)494 3973 home 40 South Alcaniz St. (850)202 4416 office Pensacola (850)202 4440 fax FL 32501 (850)291 0667 cell phayes@ihmc.us http://www.ihmc.us/users/phayes
Attachments
- application/octet-stream attachment: section6.2.tex
- text/plain attachment: Sections5.1and5.2.tex
Received on Thursday, 15 April 2004 02:53:47 UTC