Technically it is possible to demonstrate a technology that can identify the video, via watermarking, fingerprinting, sampling, etc etc, to make reasonable guess & match of the video content. Facial, voice recognition technology are also well-established that it can be reasonable adapted.
Deploying these technology in a network protocol however is a different set of problem however, particularly the computation and the data set. These technology is sufficient complex that it is not possible for the networking devices to handle. This means it would rely on a third party, somewhat centralized system to do the computation, which would work, except IETF, in general, have very challenging difficulties to reach rough consensus on a centralized system*. And you have this other nasty problem called "scaling" to deal with. I haven't follow on the technology development in the video matching field to say with confident how accurate it would be but for discussion, lets say 70-80% accuracy assuming no false positive, similar for voice recognition, language translation, spam identification etc. It is the same reasons why these technologies was never able to be fully adopted within IETF protocol. * Certain exceptions are made if it involved IANA, but even so, it is difficult. And in this particularly case, I doubt IANA is keen to host "fingerprints" of all the video content. -James Seng On Thu, Dec 4, 2008 at 7:16 AM, DePriest, Greg (NBC Universal) <[EMAIL PROTECTED]> wrote: > There are a number of companies that can demonstrate working technical > solutions that identify copyrighted content. It is not difficult. > > Item three at http://www.ugcprinciples.com/ references how content > identification technology is now being used by user-generated content > sites to respect copyrights. > > -----Original Message----- > From: James Seng [mailto:[EMAIL PROTECTED] > Sent: Wednesday, December 03, 2008 5:41 PM > To: DePriest, Greg (NBC Universal) > Cc: Richard Bennett; Nicholas Weaver; Paul Jessop; Craig Seidel; Le > Blond, Stevens; [email protected]; Arnaud Legout > Subject: Re: [alto] Paper on "Pushing BitTorrent Locality to the Limit" > > Good point. > > My view is that one can be reasonably solved technically without human > intervention and one is not. > > I love to see a working technical solution that can identify copyright > content. > > Sent from my iPhone > > On Dec 4, 2008, at 6:02 AM, "DePriest, Greg (NBC Universal)" > <[EMAIL PROTECTED] > > wrote: > >> That seems like a constructive suggestion. Thank you. >> >> I do have one question regarding policies. >> >> Why is protecting privacy a requirement and protecting copyrighted >> content a policy? >> >> -----Original Message----- >> From: Richard Bennett [mailto:[EMAIL PROTECTED] >> Sent: Wednesday, December 03, 2008 4:43 PM >> To: Nicholas Weaver >> Cc: DePriest, Greg (NBC Universal); [email protected]; Arnaud Legout; Paul >> Jessop; Craig Seidel; Le Blond, Stevens >> Subject: Re: [alto] Paper on "Pushing BitTorrent Locality to the >> Limit" >> >> It strikes me that the discovery of illegal content is a local policy >> decision. There are jurisdictions that require it and those that >> forbid >> it. Perhaps ALTO needs to support a policy option that allows content >> descriptors to be queried, blocked, or redirected in the interest of >> local laws and regulations. >> >> I don't want to spoil anybody's fun, of course. >> >> RB >> >> Nicholas Weaver wrote: >>> >>> On Dec 3, 2008, at 1:15 PM, DePriest, Greg (NBC Universal) wrote: >>> >>>> You note that "A localization service doesn't have to discriminate >>>> [between legit and illegit P2P]..." >>>> >>>> I don't understand why it wouldn't. >>>> >>>> What's the point of facilitating the illegal distribution of >> copyrighted >>>> content? >>>> >>>> And how would one justify that? >>> >>> Under the same justification that you allow BitTorrent at all: You >>> DON'T know that it is copyrighted, it could be Linux ISOs, with >>> enough >> >>> probability that you can't just block the protocol and you can't sue >>> BitTorrent Inc into submission under the Napster and related methods. >>> >>> Or that you allow HTTP traffic, after all, that could be copyrighted >>> material, kiddie porn, or other bad content. >>> >>> It is not the responsibility of the network to police content, and a >>> localization service doesn't actually have to know what it is >>> localizing, so it is not in a position to police content one way or >>> the other. >>> >>> EG, ask localization service "Who else is accessing 512b-random-ID >>> SHA-512 file descriptor", and the localization service has no notion >>> what the resource is, just a list of who's accessing it. Its in many >>> ways easier to make a localization service which is agnostic. >>> >>> _______________________________________________ >>> alto mailing list >>> [email protected] >>> https://www.ietf.org/mailman/listinfo/alto >> >> -- >> Richard Bennett >> >> _______________________________________________ >> alto mailing list >> [email protected] >> https://www.ietf.org/mailman/listinfo/alto > _______________________________________________ alto mailing list [email protected] https://www.ietf.org/mailman/listinfo/alto
