At 01:11 PM 12/11/96 -0800, you wrote:
PICS is the wrong approach becuase it oversimplifies the ratings of content, because it places the ratings made by the author in the payload itself, and because third-party ratings systems are cut out of the loop (effectively).
Um, first, there's nothing in the PICS spec that requires ratings to be embedded with the content. PICS labels can be distributed one of three ways: 1) embedded in the content, assuming there's a method of embedding defined for the particular type of content. (For HTML, you can put it in a META tag.) 2) sent "along-with" the content as part of the transmission protocol. (For HTTP, there's a standard by which PICS labels can be sent as RFC-822 headers in the HTTP reply, but no one is using that to the best of my knowledge.) 3) distributed from the third-party label bureau. Method of distribution is independent of author of the label, too. It's perfectly reasonable for the author to distribute labels for his content via a third-part label bureau, or for an author to embed a label from the GoodMouseClicking page rating service within his document. By the way, PICS labels nominally apply to a document named by a particular URL. You can elide the URL, I think, if the label is sent along with the content or embedded within the content, but when you ask a label bureau for labels you request them by URL. Second, I don't see how PICS oversimplifies content rating. If anything, I would expect complaints that PICS complicates things too much because there can be an infinite number of rating systems and a human is forced into the loop to read and evaluate the meaning of any particular rating system. It is true that PICS currently only permits rating values to be numbers, but enough PICS users need non-numeric values that I expect this to be changed in January at the PICS WG meeting.
One computerish way to think of this is that the "binding" is too early. At the time of distribution, say, I mark my work something with some PICS label, based upon my best understanding of the PICS labels, ratings, agencies, and laws. But once set, the "binding" has been made. Later reviews or reviews by other entities cannot affect the binding, at least not for this distributed instance.
It's true that reviewer B cannot affect reviewer A's labels, but B can make statements about the quality of A's labels, and I can choose (in a more general trust management environment) to accept labels only from label authors who are vouched for by some particular vouching service. So GoodMouseClicking might say I'm a reliable rater of content in "Bal's crypto-relevant rating service" but a lousy judge of pages for "Joe's cool jazz pages rating service."
More importantly, the "payload" does not carry some particular set of fairly-arbitrary PICS evluations. Binding by the censors instead of by the originator, which is as it should be.
The Feds obviously believe in "encouraging" self-rating as a means of defending yourself when they haul you into court, but in general I think people will tend to defer trust to particular third-party ratings services that they choose over an author's self-labels. After all, if I'm looking for movie reviews the last thing I read is the self-promotion put out by the distributor; I look for a third-party I know who tends to agree with my tastes. Same thing with product reviews (e.g. Consumer Reports) or book reviews. --bal