- From: Bullard, Claude L \(Len\) <len.bullard@intergraph.com>
- Date: Wed, 31 May 2006 15:20:21 -0500
- To: <noah_mendelsohn@us.ibm.com>
- Cc: "Frank Manola" <fmanola@acm.org>, <www-tag@w3.org>
Then you have measures of reasonableness for a game and the measures applied are related to the game, risk and reward. I note the introduction of "known source of" in your reply. For a 'known source of good weather forecasts" one assume knowledge a priori of the play/access. In that case, the measure of reasonableness is much less expensive. The representation doesn't convey that; the association does. I think we agree here. The trust is within the context of associations to other sources. That's fine. Trust but verify. The web is caveat emptor because any claims-based system is on a blind play (no context of associations (usually episodic and pragmatic). While it is a reasonable strategy to play 'tit for tat' and do as done unto without massive retaliation for defection, it is nonetheless a test-based commitment. That one can also use other sources of information to make it easier to make that first commitment doesn't change the nature of the system architecture. You make the commitment relative to the risk: if it is a zero sum game, you assess the loss. If it is not a zero sum game, you assess the loss relative to a series of plays. If the loss or win does not change the rules, you are in a Nash equilibrium and you assess the value of the game itself. len From: noah_mendelsohn@us.ibm.com [mailto:noah_mendelsohn@us.ibm.com] Len Bullard writes: > It is reasonable to assume nothing until using the service. First > access is blind. What the claims establish are the conditions to test > by access (is this a weather report), and in some claims, repeated > access (is this the best weather report). The proof is in the using. > > The metadata presents claims to be verified. The URI is agnostic to the > metadata claims. It is the user that has to be reasonable through > observation (use and memory of use). I'm not sure I see things quite this way. Consider the weather report example. While it's true that a user or even software can often determine that data is erroneous, e.g. because what came back was a stock quote instead of a weather report, it's atypical to be able to prove that the data is correct. I may, for example, be satisfied that the information retrieved >appears to be< a weather forecast for the intended city, but my trust that it is a correct, current and reliable forecast is likely to be based in part on just the sorts of external factors mentioned in the draft finding. So, if I see the URI listed on a seemingly current billboard for a known source of good weather forecasts, that contributes to my belief that the forecast retrieved is in fact a good one. If I get a similar web page by trying random URIs, I may note that it looks like a weather forecast, but I will trust it a lot less. So, I agree that to some extent the metadata is suspect, but I don't agree with the implication that the verification will come entirely from inspection of the retrieved representation. It's certainly desirable for representations on the Web to be self-describing, but I think that some of the trust one has in the information retrieved can be based on other representations made by the resource authority. These may be in the form of normative specifications for its URI assignment policies, or may be provided less formally (and probably less reliably) in advertisements and the like. > AFAIK, there is no architectural solution to a priori trust of > information resources. The web is a caveat emptor system by design. > Any claims-based system is. Yes, but the sources of confirming information are not in all cases limited to what is retrieved from a GET, but may come from other specifications or statements that can be, with reasonable reliability, traced to the assignment authority. -------------------------------------- Noah Mendelsohn IBM Corporation One Rogers Street Cambridge, MA 02142 1-617-693-4036 --------------------------------------
Received on Wednesday, 31 May 2006 20:20:30 UTC