Good morning all
", "time": "2025-09-30T07:15:40.000Z"}, {"author": "Brendan Quinn", "text": "I'm guessing that mouse pointer is poking around some settings panels that we can't see...
", "time": "2025-09-30T07:15:56.000Z"}, {"author": "Brendan Quinn", "text": "we can hear you but with echo
", "time": "2025-09-30T07:16:48.000Z"}, {"author": "Max Gendler", "text": "Can hear you! Light echo, Mark
", "time": "2025-09-30T07:16:54.000Z"}, {"author": "Suresh Krishnan", "text": "The room layout is a bit challenging
", "time": "2025-09-30T07:17:13.000Z"}, {"author": "Suresh Krishnan", "text": "But we will try to get the camera on the speakers if possible
", "time": "2025-09-30T07:17:29.000Z"}, {"author": "Elaine Newton", "text": "question from someone relatively new to the process: does the charter represent what the WG should consider for standardization or what it must standardize?
", "time": "2025-09-30T07:29:54.000Z"}, {"author": "Suresh Krishnan", "text": "There are two pieces Elaine
", "time": "2025-09-30T07:30:14.000Z"}, {"author": "Suresh Krishnan", "text": "The overall charter talks about background and problem space
", "time": "2025-09-30T07:30:26.000Z"}, {"author": "Suresh Krishnan", "text": "But we do have specific deliverables
", "time": "2025-09-30T07:30:33.000Z"}, {"author": "Eric Rescorla", "text": "More must than should :)
", "time": "2025-09-30T07:30:40.000Z"}, {"author": "Eric Rescorla", "text": "https://datatracker.ietf.org/wg/aipref/about/
", "time": "2025-09-30T07:33:49.000Z"}, {"author": "Mike Bishop", "text": "https://datatracker.ietf.org/wg/aipref/about/
", "time": "2025-09-30T07:33:51.000Z"}, {"author": "Martin Thomson", "text": "The Working Group will deliver:
\nA standard track document covering vocabulary for expressing AI-related preferences, independent of how those preferences are associated with content.\nStandard track document(s) describing means of attaching or associating those preferences with content in IETF-defined protocols and formats, including but not limited to using Well-Known URIs (RFC 8615) such as the Robots Exclusion Protocol (RFC 9309), and HTTP response header fields.\nA standard method for reconciling multiple expressions of preferences.\nOh wow, meetecho formatting fail.
", "time": "2025-09-30T07:35:01.000Z"}, {"author": "Jo Levy", "text": "one significant concern is that the charter is too broad, covering how content is collected. Rather than limited to use. Among other concerns
", "time": "2025-09-30T07:36:26.000Z"}, {"author": "Warren Kumari", "text": "Maybe another poll or \"norrower\" vs \"Wider\"?
", "time": "2025-09-30T07:38:58.000Z"}, {"author": "Erin Simon", "text": "agreed
", "time": "2025-09-30T07:39:14.000Z"}, {"author": "Warren Kumari", "text": "We could spend an infinite amount of time nitpicking on the charter wording....
", "time": "2025-09-30T07:40:00.000Z"}, {"author": "Felix Reda", "text": "You're welcome to participate in minutes, I'm not able to catch all that is said: https://notes.ietf.org/Vc4ezWRkTMq2xcGF4MecWg?edit
", "time": "2025-09-30T07:42:55.000Z"}, {"author": "Leonard Rosenthol", "text": "Bullet #1 is clear that the vocab is \"independent of how those preferences are associated with content\"
", "time": "2025-09-30T07:44:37.000Z"}, {"author": "Martin Thomson", "text": "There is no mention of a distinction between offline and online. I'm increasingly of the opinion that this distinction is not a useful one.
", "time": "2025-09-30T07:44:59.000Z"}, {"author": "Erin Simon", "text": "how the vocabulary is associated to particular content is not the same as what activities the vocabulary applies to
", "time": "2025-09-30T07:45:54.000Z"}, {"author": "Leonard Rosenthol", "text": "+1 to Martin. The distinction is about the attachment methods. (admittedly some may work in offline scenarios, but that is secondary)
", "time": "2025-09-30T07:46:02.000Z"}, {"author": "Max Gendler", "text": "As an aside, can someone link to the minutes, struggling to find them through the agenda page for some reason.
", "time": "2025-09-30T07:46:47.000Z"}, {"author": "Irina Bejan", "text": "https://notes.ietf.org/Vc4ezWRkTMq2xcGF4MecWg?edit
", "time": "2025-09-30T07:46:58.000Z"}, {"author": "Martin Thomson", "text": "There's a link to minutes at the top of meetecho
", "time": "2025-09-30T07:46:59.000Z"}, {"author": "Warren Kumari", "text": "https://notes.ietf.org/Vc4ezWRkTMq2xcGF4MecWg?edit
", "time": "2025-09-30T07:47:05.000Z"}, {"author": "Martin Thomson", "text": "Except that the link in meetecho doesn't work
", "time": "2025-09-30T07:47:21.000Z"}, {"author": "Martin Thomson", "text": "ignore my suggestion
", "time": "2025-09-30T07:47:26.000Z"}, {"author": "Farzaneh", "text": "well we can't blame marketing folks for using AI to explain everything ... ;)
", "time": "2025-09-30T07:48:10.000Z"}, {"author": "Eric Rescorla", "text": "In fairness, I think I proposed we developed that kind of vocabulary. Regrettably, I have not actually done so :)
", "time": "2025-09-30T07:48:47.000Z"}, {"author": "Kevin Kelley", "text": "There's also Krishna's proposal, which has some elements of that, too.
", "time": "2025-09-30T07:48:59.000Z"}, {"author": "Eric Rescorla", "text": "Yes, Krishna's is more fleshed out
", "time": "2025-09-30T07:49:10.000Z"}, {"author": "Farzaneh", "text": "yes
", "time": "2025-09-30T07:49:22.000Z"}, {"author": "Kevin Kelley", "text": "This has come up before, I think, if I'm understanding the concern! (\"capability\" vs. \"use\")
", "time": "2025-09-30T07:50:24.000Z"}, {"author": "Leonard Rosenthol", "text": "however, Krishna's proposal does not address the entire charter by limiting things to only SOME forms of attachments...
", "time": "2025-09-30T07:50:42.000Z"}, {"author": "Eric Rescorla", "text": "I'm going to say this later, but my concern with the use of the term \"training\" is that it's overbroad and I think we're going to end up prohibiting a lot of useful stuff that the sites aren't really that interested in preventing
", "time": "2025-09-30T07:50:53.000Z"}, {"author": "Leonard Rosenthol", "text": "@eric - it's not just about sites, which is where I think we keep running into walls...
", "time": "2025-09-30T07:52:49.000Z"}, {"author": "Farzaneh", "text": "I think it will even get broader when we say AI training and I believe it should be narrower . and Mark just said maybe we can discuss specific purpose AI
", "time": "2025-09-30T07:52:56.000Z"}, {"author": "Martin Thomson", "text": "Leonard, it is always possible to deliver parts of a charter. That is, less than the whole. Now, you can argue that delivering part is harmful in some ways (that is possible).
", "time": "2025-09-30T07:53:43.000Z"}, {"author": "Max Gendler", "text": "@Eric Rescorla I think another issue with overbroad categories is also that it doesn't allow nuance in preferences, which leads to difficulties for content owners
", "time": "2025-09-30T07:53:50.000Z"}, {"author": "Warren Kumari", "text": "\"What problem are we trying to solve...\"
", "time": "2025-09-30T07:54:16.000Z"}, {"author": "Fabrice Canel", "text": "Remote participants are experiencing echo, while in-room speakers are unaffected. Mark and Suresh, please consider muting your microphones when remote speakers are talking. Thanks.
", "time": "2025-09-30T07:54:31.000Z"}, {"author": "Farzaneh", "text": "I still have trouble with using \"content owners\" and publishers. they are not always the ones who signal the preferences ... we assume ownership if we keep saying it
", "time": "2025-09-30T07:54:55.000Z"}, {"author": "Suresh Krishnan", "text": "Will check on it Fabrice
", "time": "2025-09-30T07:55:09.000Z"}, {"author": "Max Gendler", "text": "Farzaneh said:
\n\n\nI still have trouble with using \"content owners\" and publishers. they are not always the ones who signal the preferences ... we assume ownership if we keep saying it
\n
Can you expand on who else should be signaling preferences that does not fit into those two categories?
", "time": "2025-09-30T07:55:46.000Z"}, {"author": "Suresh Krishnan", "text": "@Fabrice, Mark's mic is now muted
", "time": "2025-09-30T07:56:07.000Z"}, {"author": "Suresh Krishnan", "text": "Did that help with the echo?
", "time": "2025-09-30T07:56:12.000Z"}, {"author": "Farzaneh", "text": "sometimes the declarant (as Paul puts it in the ID) is not the content owner. nobody owns everything on their website.
", "time": "2025-09-30T07:56:57.000Z"}, {"author": "Max Gendler", "text": "Farzaneh said:
\n\n\nsometimes the declarant (as Paul puts it in the ID) is not the content owner. nobody owns everything on their website.
\n
I would strongly disagree with the idea that 'nobody owns everything on their website\" - but your answer doesn't actually tell me who a declarant could be that isn't a content owner or the publisher of the content. What third category exists for you?
", "time": "2025-09-30T07:58:32.000Z"}, {"author": "Fabrice Canel", "text": "Yes Suresh .... External speaker echoing was gone. Thank you.
", "time": "2025-09-30T07:59:28.000Z"}, {"author": "Suresh Krishnan", "text": "Great Fabrice
", "time": "2025-09-30T08:00:04.000Z"}, {"author": "Max Gendler", "text": "Echo is back with Leonard for me unfortunately
", "time": "2025-09-30T08:00:25.000Z"}, {"author": "Brendan Quinn", "text": "External speaker echo is back - maybe it depends on the speaker somehow?
", "time": "2025-09-30T08:00:28.000Z"}, {"author": "Suresh Krishnan", "text": "Speaker is remote now
", "time": "2025-09-30T08:00:39.000Z"}, {"author": "Suresh Krishnan", "text": "As far as I can see there are no room mics on at this point
", "time": "2025-09-30T08:01:01.000Z"}, {"author": "Martin Thomson", "text": "There is one mic always on: the one in front of the podium.
", "time": "2025-09-30T08:01:21.000Z"}, {"author": "Farzaneh", "text": "site operators Max and I believe there are many more categories that we can't think of. also it depends on what you mean by \"owning\" does Wikipedia own everything on its site? or does the community that create it owns them
", "time": "2025-09-30T08:01:53.000Z"}, {"author": "Martin Thomson", "text": "Why does this have to be about \"ownership\" or \"rights\"?
", "time": "2025-09-30T08:03:27.000Z"}, {"author": "Bradley Silver", "text": "Ownership is not a predicate to having an interest sufficient to warrant expressing a preference.
", "time": "2025-09-30T08:03:33.000Z"}, {"author": "Erin Simon", "text": "a preference that others are free to disregard, then
", "time": "2025-09-30T08:04:01.000Z"}, {"author": "Bradley Silver", "text": "You don't need to be an owner of anything to use robots.txt for example
", "time": "2025-09-30T08:04:25.000Z"}, {"author": "Martin Thomson", "text": "what it means for different entities to disregard preferences seems to be an important part of the question, yes.
", "time": "2025-09-30T08:04:31.000Z"}, {"author": "Kevin Kelley", "text": "Brad: I'd argue that that's one of the issues with the way that robots.txt is being used today.
", "time": "2025-09-30T08:05:22.000Z"}, {"author": "Warren Kumari", "text": "\"provided it is productive\" :-P
", "time": "2025-09-30T08:05:36.000Z"}, {"author": "Warren Kumari", "text": "I'll note I love snark....
", "time": "2025-09-30T08:05:46.000Z"}, {"author": "Eric Rescorla", "text": "A year is lightning fast for IETF
", "time": "2025-09-30T08:06:12.000Z"}, {"author": "Suresh Krishnan", "text": "100% @ekr
", "time": "2025-09-30T08:06:24.000Z"}, {"author": "Martin Thomson", "text": "Work can be done in less than a year, but that is highly abnormal
", "time": "2025-09-30T08:06:41.000Z"}, {"author": "Eric Rescorla", "text": "Vote early vote often
", "time": "2025-09-30T08:07:24.000Z"}, {"author": "Eric Rescorla", "text": "In lexical sort order of the SHA-3 hash of your name
", "time": "2025-09-30T08:12:34.000Z"}, {"author": "Jacob Hoffman-Andrews", "text": "Jacob Hoffman-Andrews - Public Interest Technologist at EFF :wave:
", "time": "2025-09-30T08:13:50.000Z"}, {"author": "Suresh Krishnan", "text": "Back in 19 minutes
", "time": "2025-09-30T08:16:49.000Z"}, {"author": "Suresh Krishnan", "text": "We are starting
", "time": "2025-09-30T08:40:44.000Z"}, {"author": "Kevin Kelley", "text": "Isn't that outcome that no preference is expressed?
", "time": "2025-09-30T08:48:33.000Z"}, {"author": "Kevin Kelley", "text": "I'm not sure that I understand your concern around defaults, Max. Could you say more?
", "time": "2025-09-30T08:48:54.000Z"}, {"author": "Max Gendler", "text": "FWIW could hear Kevin and EKR well without mic.
\nKevin, sorry if I'm being unclear, it's 4:50 in the morning here and I'm trying my best :)
\nI think if we create a limited vocabulary, the defaults are permissive for anything that is undefined, which I think creates ecosystem issues. Is that clearer?
No worries, I'm pretty jet-lagged myself! That does make more sense, yes, thanks.
", "time": "2025-09-30T08:50:58.000Z"}, {"author": "Eric Rescorla", "text": "Max, btw, I was using the mic
", "time": "2025-09-30T08:51:25.000Z"}, {"author": "Kevin Kelley", "text": "Do you see that as being different or worse than the status quo that we're all here to work on?
", "time": "2025-09-30T08:51:34.000Z"}, {"author": "Max Gendler", "text": "Eric Rescorla said:
\n\n\nMax, btw, I was using the mic
\n
Before you picked it up I heard you well is what I meant. For both.
", "time": "2025-09-30T08:52:08.000Z"}, {"author": "Eric Rescorla", "text": "Max, thanks
", "time": "2025-09-30T08:52:14.000Z"}, {"author": "Max Gendler", "text": "Kevin Kelley said:
\n\n\nDo you see that as being different or worse than the status quo that we're all here to work on?
\n
\"Same same, but different\" - similar problem in some ways, different in some. Worse and better than status quo in some ways and I'd love to just get us all to better than status quo and I don't think it's out of the question. That's not a particularly helpful answer, but the helpful answer takes a longer time than I'd like to dump in chat
", "time": "2025-09-30T08:53:25.000Z"}, {"author": "Eric Rescorla", "text": "I'd like to hear some more about why people are concerned about training specifically. Because when see people explain why they object to training, it usually comes down to some more concrete objection. I agree it's not just about substitution
", "time": "2025-09-30T08:53:29.000Z"}, {"author": "Kevin Kelley", "text": "Sure---thanks, Max. I'd love to hear the helpful answer at some point, though :)
", "time": "2025-09-30T08:56:28.000Z"}, {"author": "Elaine Newton", "text": "If mic muting would help the echo, it would be greatly appreciated from out here in the ether.
", "time": "2025-09-30T08:57:04.000Z"}, {"author": "Max Gendler", "text": "Kevin Kelley said:
\n\n\nSure---thanks, Max. I'd love to hear the helpful answer at some point, though :)
\n
Happy to, I'm assuming it will come up throughout the next couple of days in conversation, if not we can loop back.
", "time": "2025-09-30T08:57:16.000Z"}, {"author": "Suresh Krishnan", "text": "@Elaine: we are trying to mute the room mic other than the speakers
", "time": "2025-09-30T08:58:04.000Z"}, {"author": "Suresh Krishnan", "text": "It is harder when the speaker is remote
", "time": "2025-09-30T08:58:12.000Z"}, {"author": "Kevin Kelley", "text": "Re: Pedro: I think that another reason to stay slightly higher-level is this technology and its uses are evolving rapidly---I'm not sure that anyone in the room knows what this will look like in 6 or 12 months.
", "time": "2025-09-30T08:58:38.000Z"}, {"author": "Eric Rescorla", "text": "Usually these hand mics have a very short range, so I suspect it's the room mics
", "time": "2025-09-30T08:58:54.000Z"}, {"author": "Jacob Hoffman-Andrews", "text": "According to Meetecho the extra sound is coming from \"Mark Nottingham\" which is presumably the room mic.
", "time": "2025-09-30T08:59:44.000Z"}, {"author": "Caleb Donaldson", "text": "aptly names meet echo
", "time": "2025-09-30T09:02:11.000Z"}, {"author": "Caleb Donaldson", "text": "*named
", "time": "2025-09-30T09:02:18.000Z"}, {"author": "Eric Rescorla", "text": "Jacob Hoffman-Andrews said:
\n\n\nAccording to Meetecho the extra sound is coming from \"Mark Nottingham\" which is presumably the room mic.
\n
Actually
", "time": "2025-09-30T09:02:20.000Z"}, {"author": "Eric Rescorla", "text": "Actually Mark is just repeating everything everyone says
", "time": "2025-09-30T09:02:27.000Z"}, {"author": "Eric Rescorla", "text": "2C2PA
", "time": "2025-09-30T09:03:06.000Z"}, {"author": "Martin Thomson", "text": "One 2 is enough
", "time": "2025-09-30T09:03:20.000Z"}, {"author": "Max Gendler", "text": "2C2PA2Furious
", "time": "2025-09-30T09:03:32.000Z"}, {"author": "Martin Thomson", "text": "The 2 in C2PA means that you get two copies anyway. Don't add more or you'll gremlins it.
", "time": "2025-09-30T09:03:51.000Z"}, {"author": "Leonard Rosenthol", "text": "I think one of those was supposed to be CAWG
", "time": "2025-09-30T09:04:24.000Z"}, {"author": "Eric Rescorla", "text": "C4PA
", "time": "2025-09-30T09:04:24.000Z"}, {"author": "Martin Thomson", "text": "ai-input was the inference/ai-use we talked about mostly
", "time": "2025-09-30T09:04:56.000Z"}, {"author": "Max Gendler", "text": "C22PA
", "time": "2025-09-30T09:05:24.000Z"}, {"author": "Martin Thomson", "text": "I tried to get to the reason from RSL github history, but it's empty.
", "time": "2025-09-30T09:06:04.000Z"}, {"author": "Martin Thomson", "text": "This is arguably also not a reference to extensions to RFC 9309
", "time": "2025-09-30T09:08:40.000Z"}, {"author": "Martin Thomson", "text": "I am not a lawyer, so season accordingly
", "time": "2025-09-30T09:08:52.000Z"}, {"author": "Suresh Krishnan", "text": "I also read it the same way Martin
", "time": "2025-09-30T09:09:17.000Z"}, {"author": "Suresh Krishnan", "text": "Also not a lawyer
", "time": "2025-09-30T09:09:43.000Z"}, {"author": "Max Gendler", "text": "I think it's naive to suggest that the vocabulary is not connected to attachments. They're not the same, but they're not independent.
\nFWIW, also not a lawyer, but I read it as extensions will be referenced here.
", "time": "2025-09-30T09:09:45.000Z"}, {"author": "Erin Simon", "text": "EU COP doesn't only make mandatory extensions of the robots.txt protocol (note: the protocol, not just the RFC), but it also makes mandatory other state-of-the-art mechanisms that are adopted by general agreement among rightsholders and ai providers
", "time": "2025-09-30T09:10:02.000Z"}, {"author": "Erin Simon", "text": "so it's unfortunately not the case that this group can escape legal consequences via alternate attachment mechanisms
", "time": "2025-09-30T09:10:23.000Z"}, {"author": "Kevin Kelley", "text": "Right, Erin is referring to part (b), which it's interesting was left off the slide.
", "time": "2025-09-30T09:10:30.000Z"}, {"author": "Eric Rescorla", "text": "Erin: to be clear, you're saying that one is subject to the union of robots.txt and state-of-the-art mechanisms
", "time": "2025-09-30T09:10:45.000Z"}, {"author": "Suresh Krishnan", "text": "Thanks Erin!
", "time": "2025-09-30T09:10:57.000Z"}, {"author": "Erin Simon", "text": "yes
", "time": "2025-09-30T09:11:03.000Z"}, {"author": "Erin Simon", "text": "they are two alternative pathways to arrive at a mandatory opt out
", "time": "2025-09-30T09:11:14.000Z"}, {"author": "Eric Rescorla", "text": "TY!
", "time": "2025-09-30T09:11:18.000Z"}, {"author": "Erin Simon", "text": "yw :)
", "time": "2025-09-30T09:11:42.000Z"}, {"author": "Kevin Kelley", "text": "Hm, interesting: I don't read it that way.
", "time": "2025-09-30T09:11:46.000Z"}, {"author": "Kevin Kelley", "text": "(Then again, I'm not a lawyer.)
", "time": "2025-09-30T09:11:52.000Z"}, {"author": "Martin Thomson", "text": "b) to identify and comply with other appropriate machine-readable protocols to express
\nrights reservations pursuant to Article 4(3) of Directive (EU) 2019/790, for example
\nthrough asset-based or location-based metadata, that have either have been adopted
\nby international or European standardisation organisations, or are state-of-the-art,
\nincluding technically implementable, and widely adopted by rightsholders,
\nconsidering different cultural sectors, and generally agreed through an inclusive
\nprocess based on bona fide discussions to be facilitated at EU level with the
\ninvolvement of rightsholders, AI providers and other relevant stakeholders as a more
\nimmediate solution, while anticipating the development of standards.
That seems separable to the point (a) about robots.txt
", "time": "2025-09-30T09:12:51.000Z"}, {"author": "Leonard Rosenthol", "text": "+1 to Felix! Attachments requires Vocab...
", "time": "2025-09-30T09:13:19.000Z"}, {"author": "Eric Rescorla", "text": "I would not want to have to be the person whose job it was to determine whether a given mechanism was \"are state-of-the-art, including technically implementable, and widely adopted by rightsholders\"
", "time": "2025-09-30T09:13:33.000Z"}, {"author": "Erin Simon", "text": "it's not such a bad job, as jobs go
", "time": "2025-09-30T09:14:32.000Z"}, {"author": "Martin Thomson", "text": "Success (or not) is determined, not in process, but in the minds of every person in the room on their own. We do not seek to assess consensus about whether something is successful.
", "time": "2025-09-30T09:14:44.000Z"}, {"author": "Suresh Krishnan", "text": "@Paul: In the IETF sense a successor to RFC9309 would Obsolete it.
", "time": "2025-09-30T09:14:46.000Z"}, {"author": "Eric Rescorla", "text": "@Suresh Krishnan or UPDATE
", "time": "2025-09-30T09:15:13.000Z"}, {"author": "Eric Rescorla", "text": "updates
", "time": "2025-09-30T09:15:38.000Z"}, {"author": "Suresh Krishnan", "text": "Yep @ekr. But it might be more subjective since Update is a bit more vague
", "time": "2025-09-30T09:15:41.000Z"}, {"author": "Martin Thomson", "text": "For those who are not up on the IETF history, documents can \"update\" others, but the effect of an \"update\" is not well understood. For the question of how that applies to the text Paul showed, the answer is a big fat \"who knows\".
", "time": "2025-09-30T09:17:14.000Z"}, {"author": "Eric Rescorla", "text": "Sure. I actually think that the tags are not really the important thing here. What's relevant is what the document says about itself.
", "time": "2025-09-30T09:18:33.000Z"}, {"author": "Eric Rescorla", "text": "Like if it calls itself Robots.txt 2.0
", "time": "2025-09-30T09:18:43.000Z"}, {"author": "Eric Rescorla", "text": "Whatever tags it uses
", "time": "2025-09-30T09:18:49.000Z"}, {"author": "Suresh Krishnan", "text": "Yep. Like Martin said it is a bit fuzzy. Mirja and I wrote up something in this regard
", "time": "2025-09-30T09:19:05.000Z"}, {"author": "Suresh Krishnan", "text": "So if you want a totally unofficial primer. https://datatracker.ietf.org/doc/html/draft-kuehlewind-rswg-updates-tag-02#name-introduction
", "time": "2025-09-30T09:19:19.000Z"}, {"author": "Martin Thomson", "text": "If the IETF achieves consensus about a robots.txt 2.0, then it will replace RFC 9309. That's true. It would likely obsolete RFC 9309.
", "time": "2025-09-30T09:19:27.000Z"}, {"author": "Thom Vaughan", "text": "EKR++ 100% agreed
", "time": "2025-09-30T09:19:39.000Z"}, {"author": "Elaine Newton", "text": "well said
", "time": "2025-09-30T09:19:43.000Z"}, {"author": "Farzaneh", "text": "Agreed!
", "time": "2025-09-30T09:19:48.000Z"}, {"author": "Erin Simon", "text": "+1
", "time": "2025-09-30T09:19:48.000Z"}, {"author": "Eric Rescorla", "text": "100% agree to what Kevin just said about attachment
", "time": "2025-09-30T09:21:04.000Z"}, {"author": "Martin Thomson", "text": "I do have slides prepared for attachment, but the expectation here is that we won't get to that this week unless we have a significant breakthrough on issues.
", "time": "2025-09-30T09:21:56.000Z"}, {"author": "Warren Kumari", "text": "What happens to all of the deployed signals?
", "time": "2025-09-30T09:22:23.000Z"}, {"author": "Eric Rescorla", "text": "Martin Thomson said:
\n\n\nIf the IETF achieves consensus about a robots.txt 2.0, then it will replace RFC 9309. That's true. It would likely obsolete RFC 9309.
\n
True. OTOH, would you say that HTTP/3 is a successor to HTTP/2? It doesn't obsolete it
", "time": "2025-09-30T09:22:46.000Z"}, {"author": "Kevin Kelley", "text": "That's a big concern of mine, Warren. I also noticed that a lot of these draft pseudo-standards don't make much allowance for versioning or forward compatibility.
", "time": "2025-09-30T09:22:49.000Z"}, {"author": "Kevin Kelley", "text": "And once things are out there on the internet...
", "time": "2025-09-30T09:23:00.000Z"}, {"author": "Warren Kumari", "text": "Yup....
", "time": "2025-09-30T09:23:04.000Z"}, {"author": "Eric Rescorla", "text": "Kevin Kelley said:
\n\n\nThat's a big concern of mine, Warren. I also noticed that a lot of these draft pseudo-standards don't make much allowance for versioning or forward compatibility.
\n
This is actually true for us as well. We might considere versioning our names so we can change the semantics
", "time": "2025-09-30T09:23:29.000Z"}, {"author": "Mirja K\u00fchlewind", "text": "Warren Kumari said:
\n\n\nWhat happens to all of the deployed signals?
\n
crawlers may or may not use them, similar as the new signals....
", "time": "2025-09-30T09:23:32.000Z"}, {"author": "Max Gendler", "text": "As a complete sidenote - Thursday is Yom Kippur, so I won't be around to talk about attachments since that's the only time we've set aside for it in particular.
", "time": "2025-09-30T09:23:49.000Z"}, {"author": "Eric Rescorla", "text": "Mirja K\u00fchlewind said:
\n\n\nWarren Kumari said:
\n\n\nWhat happens to all of the deployed signals?
\ncrawlers may or may not use them, similar as the new signals....
\n
But if they use the same names....
", "time": "2025-09-30T09:23:51.000Z"}, {"author": "Warren Kumari", "text": "Or if I beleive that I have signaled X, and the standard doesn't support X, then I have a false sense of what will happen.
", "time": "2025-09-30T09:24:41.000Z"}, {"author": "Warren Kumari", "text": "Users shouldn't need to be following every step of this process...
", "time": "2025-09-30T09:25:29.000Z"}, {"author": "Eric Rescorla", "text": "Well, we are in Europe :)
", "time": "2025-09-30T09:25:31.000Z"}, {"author": "Martin Thomson", "text": "Eric Rescorla said:
\nThis is actually true for us as well. We might considere versioning our names so we can change the semantics
We are versioning things, following the guidance in RFC 9170
", "time": "2025-09-30T09:25:38.000Z"}, {"author": "Mirja K\u00fchlewind", "text": "there will always be a certain mismatch between what a signal provider will expect and what the signal user will understand because there is no sharp definition
", "time": "2025-09-30T09:25:59.000Z"}, {"author": "Kevin Kelley", "text": "Re: ekr@: I have ideas for doing that (which I know I keep saying that I'll share more widely; sorry)---it just feels like putting the cart ahead of the horse a little bit.
", "time": "2025-09-30T09:26:45.000Z"}, {"author": "Martin Thomson", "text": "Ugh, meetecho can't handle edits to comments that come through zulip
", "time": "2025-09-30T09:26:53.000Z"}, {"author": "Max Gendler", "text": "Mirja K\u00fchlewind said:
\n\n\nthere will always be a certain mismatch between what a signal provider will expect and what the signal user will understand because there is no sharp definition
\n
Isn't that the core of what we're trying to fix? Ambiguity rarely favors what the signal provider is trying to express.
", "time": "2025-09-30T09:26:59.000Z"}, {"author": "Felix Reda", "text": "Stefano, could you take a look if I faithfully captured your intervenion in the minutes? https://notes.ietf.org/Vc4ezWRkTMq2xcGF4MecWg?edit
", "time": "2025-09-30T09:27:17.000Z"}, {"author": "Martin Thomson", "text": "I meant to say that the versioning question is addressed already, following RFC 9170
", "time": "2025-09-30T09:27:18.000Z"}, {"author": "Stefano Gentile", "text": "@Felix: sure
", "time": "2025-09-30T09:27:34.000Z"}, {"author": "Mirja K\u00fchlewind", "text": "at least we try to improve; but I don't think we can absolutely fix it.
", "time": "2025-09-30T09:27:39.000Z"}, {"author": "Gary Illyes", "text": "@martin : we also need the signals to be usable though. We have problems with people misspelling \"disallow\", imagine what happens if we start versioning verbs
", "time": "2025-09-30T09:27:43.000Z"}, {"author": "Eric Rescorla", "text": "Gary Illyes said:
\n\n\n@martin : we also need the signals to be usable though. We have problems with people misspelling \"disallow\", imagine what happens if we start versioning verbs
\n
Note that the web can't even spell \"referrer\"
", "time": "2025-09-30T09:28:00.000Z"}, {"author": "Suresh Krishnan", "text": "If the text is blurry, you can follow along at
", "time": "2025-09-30T09:28:06.000Z"}, {"author": "Suresh Krishnan", "text": "https://github.com/ietf-wg-aipref/drafts/issues
", "time": "2025-09-30T09:28:06.000Z"}, {"author": "Martin Thomson", "text": "Sure, but the way to address that is not to put a version signal on the thing.
", "time": "2025-09-30T09:28:07.000Z"}, {"author": "Martin Thomson", "text": "Suresh Krishnan said:
\n\n\nIf the text is blurry, you can follow along at
\n
That is a very poor use of 400kbps.
", "time": "2025-09-30T09:29:14.000Z"}, {"author": "Kevin Kelley", "text": "It's also very distinct from the idea of expressing rightsholder preferences.
", "time": "2025-09-30T09:30:35.000Z"}, {"author": "Martin Thomson", "text": "By \"floating point\", Ekr is talking about model weights
", "time": "2025-09-30T09:31:54.000Z"}, {"author": "Eric Rescorla", "text": "Martin Thomson said:
\n\n\nBy \"floating point\", Ekr is talking about model weights
\n
Indeed. That's metonymy
", "time": "2025-09-30T09:35:11.000Z"}, {"author": "Kevin Kelley", "text": "ekr@: That's not how I read RFC 9309.
", "time": "2025-09-30T09:36:40.000Z"}, {"author": "Eric Rescorla", "text": "Kevin Kelley said:
\n\n\nekr@: That's not how I read RFC 9309.
\n
By \"That\" you mean \"Just don't scrape me\"?
", "time": "2025-09-30T09:37:00.000Z"}, {"author": "Eric Rescorla", "text": "Is an MLM a Medium Language Model?
", "time": "2025-09-30T09:37:11.000Z"}, {"author": "Martin Thomson", "text": "Multi-Level Marketing.
", "time": "2025-09-30T09:37:26.000Z"}, {"author": "Kevin Kelley", "text": "ekr@: \"It may be inconvenient for service owners...\" (RFC 9309, \u00a71).
", "time": "2025-09-30T09:37:46.000Z"}, {"author": "Stefano Gentile", "text": "@Felix, the text looks good to me. I have shared some ley lines in pvt for good measure
", "time": "2025-09-30T09:38:04.000Z"}, {"author": "Kevin Kelley", "text": "RFC 9309 is a mechanism for service owners to direct client traffic. It's sort of wound up being used for other things because it's the best-available option, not because it's a good option or because it's a desirable one.
", "time": "2025-09-30T09:38:21.000Z"}, {"author": "Martin Thomson", "text": "Yeah, robots.txt gets used because that's where the crawler is looking already.
", "time": "2025-09-30T09:38:47.000Z"}, {"author": "Eric Rescorla", "text": "I'd like to suggest that one way to help make some progress would be see a presentation from Krishna about his alternative design. Not for the purpose of having an adoption call but to help building understanding.
", "time": "2025-09-30T09:40:09.000Z"}, {"author": "Eric Rescorla", "text": "Well, this negotiation position is based on saying \"no\"
", "time": "2025-09-30T09:40:22.000Z"}, {"author": "Farzaneh", "text": "yes I would like Krishna's ID to be presented
", "time": "2025-09-30T09:40:34.000Z"}, {"author": "Eric Rescorla", "text": "Kevin Kelley said:
\n\n\nRFC 9309 is a mechanism for service owners to direct client traffic. It's sort of wound up being used for other things because it's the best-available option, not because it's a good option or because it's a desirable one.
\n
I don't necessarily disagree with that.
", "time": "2025-09-30T09:40:41.000Z"}, {"author": "Kevin Kelley", "text": "Re: Michael---this was the \"narrow\" vs \"broad\" construction thing, and the group there had a fairly solid consensus about this approach, if I recall.
", "time": "2025-09-30T09:41:30.000Z"}, {"author": "Suresh Krishnan", "text": "@ekr, @farz yes we should be able to find some time today to discuss Krishna's draft
", "time": "2025-09-30T09:41:56.000Z"}, {"author": "Martin Thomson", "text": "The challenge with this is that training is often a terminal category in the sense that there are no obvious categories that are more narrow. We've identified generative AI as distinct from the general case, but how would you split things further?
", "time": "2025-09-30T09:42:05.000Z"}, {"author": "Mike Bishop", "text": "Training is producing a tool to perform some task. I think the proposal is to describe the task, not the tool.
", "time": "2025-09-30T09:42:44.000Z"}, {"author": "Martin Thomson", "text": "ai-use is obviously very much more complex and might allow for more options underneath, if we had that. As Mike says.
", "time": "2025-09-30T09:42:55.000Z"}, {"author": "Bradley Silver", "text": "But once the training has happened, the model could theoretically be used for anything. Practically, the best means of having some say over this is a broader training preferences.
", "time": "2025-09-30T09:43:36.000Z"}, {"author": "Chris Needham", "text": "and some organisations will bundle AI usage with other usage when crawling, so it's not enough to use today's robots.txt by itself
", "time": "2025-09-30T09:44:09.000Z"}, {"author": "Kevin Kelley", "text": "Brad: that's not the first time that that idea has come up; but I'm not sure that I completely understand the objection.
", "time": "2025-09-30T09:44:22.000Z"}, {"author": "Eric Rescorla", "text": "Mike Bishop said:
\n\n\nTraining is producing a tool to perform some task. I think the proposal is to describe the task, not the tool.
\n
So as I understood what Pedro was saying earlier, you would often train the same model for multiple uses and then fine-tune it for specific applications.
", "time": "2025-09-30T09:44:27.000Z"}, {"author": "Eric Rescorla", "text": "Eric Rescorla said:
\n\n\nMike Bishop said:
\n\n\nTraining is producing a tool to perform some task. I think the proposal is to describe the task, not the tool.
\nSo as I understood what Pedro was saying earlier, you would often train the same model for multiple uses and then fine-tune it for specific applications.
\n
Which I guess he just said
", "time": "2025-09-30T09:44:39.000Z"}, {"author": "Martin Thomson", "text": "This is a common approach for LLMs, yes
", "time": "2025-09-30T09:44:53.000Z"}, {"author": "Bradley Silver", "text": "@Kevin, there is simply no confidence that saying yes to training for a specific X purpose, will reasonably be limited to that purpose.
", "time": "2025-09-30T09:45:19.000Z"}, {"author": "Farzaneh", "text": "I don\u2019t think it's appropriate to design an IETF standard enforces copyright. That\u2019s not what this process is for. The goal should be to create a protocol that supports interoperability, transparency, and user agency, not one that reinforces specific business models or limits how information can be discovered. This emphasis on content owners and \"rightsholders\" here keeps getting us to the direction of enforcement of rights through the protocol.
", "time": "2025-09-30T09:45:21.000Z"}, {"author": "Caleb Donaldson", "text": "+1
", "time": "2025-09-30T09:45:52.000Z"}, {"author": "Bradley Silver", "text": "@Pedro, search is developing in a way that either ignore or extremely limits the choices of content providers. That is important context which hasn't been mentioned...
", "time": "2025-09-30T09:46:58.000Z"}, {"author": "Felix Reda", "text": "Just FYI: None of what you're writing in chat is captured in the minutes. If you would like it to be included, please copy/paste it over there: https://notes.ietf.org/Vc4ezWRkTMq2xcGF4MecWg?edit
", "time": "2025-09-30T09:48:32.000Z"}, {"author": "Eric Rescorla", "text": "If we don't plan to let people express preferences, we could just tour around Zurich for a few days
", "time": "2025-09-30T09:48:33.000Z"}, {"author": "Pedro Suarez", "text": "@Bradley, yes I agree with that. What I'm saying is that the technology behind \"classical search\" and \"AI Overviews\" is effectively the same today. The difference is how the results are presented by the search provider
", "time": "2025-09-30T09:48:44.000Z"}, {"author": "Kevin Kelley", "text": "@Brad: Why is that objection specific to trained models?
", "time": "2025-09-30T09:48:44.000Z"}, {"author": "Kevin Kelley", "text": "(Sorry, trying to figure out how to frame my question well.)
", "time": "2025-09-30T09:49:06.000Z"}, {"author": "Farzaneh", "text": "I dont think we said people shouldnt express preferences. we disagree on how and what etc
", "time": "2025-09-30T09:49:31.000Z"}, {"author": "Martin Thomson", "text": "@Farz I continue to hear people claim that category X is too broad. I don't know it means to say that in this context.
", "time": "2025-09-30T09:50:51.000Z"}, {"author": "Kevin Kelley", "text": "If data is collected (in a world where we've shipped some of this work), the folks collecting it will need to track preference signals and use them as part of ensuring that data is used in acceptable ways. Why are you more reticent to trust that that would happen once a model is trained on some part of that data?
", "time": "2025-09-30T09:52:04.000Z"}, {"author": "Martin Thomson", "text": "I heard something more specific from Ekr, which is that a preference does not accidentally or unintentionally cover things that were not intended
", "time": "2025-09-30T09:52:05.000Z"}, {"author": "Eric Rescorla", "text": "And by sites I mean \"declarants\"
", "time": "2025-09-30T09:52:18.000Z"}, {"author": "Eric Rescorla", "text": "Martin Thomson said:
\n\n\nI heard something more specific from Ekr, which is that a preference does not accidentally or unintentionally cover things that were not intended
\n
Precisely.
", "time": "2025-09-30T09:52:30.000Z"}, {"author": "Eric Rescorla", "text": "Eric Rescorla said:
\n\n\nMartin Thomson said:
\n\n\nI heard something more specific from Ekr, which is that a preference does not accidentally or unintentionally cover things that were not intended
\nPrecisely.
\n
And that declarants not be put in the position where they have to forbid either A | B or neither.
", "time": "2025-09-30T09:53:25.000Z"}, {"author": "Martin Thomson", "text": "Understanding where lines in vocabulary align with lines in products might be helpful. I would be supportive of seeing that.
", "time": "2025-09-30T09:53:38.000Z"}, {"author": "Irina Bejan", "text": "+1
", "time": "2025-09-30T09:53:47.000Z"}, {"author": "Eric Rescorla", "text": "When what they really want to do is forbid A
", "time": "2025-09-30T09:53:47.000Z"}, {"author": "Martin Thomson", "text": "Eric Rescorla said:
\n\n\nAnd that declarants not be put in the position where they have to forbid either A | B or neither.
\n
This is a line where the \"search\" definition comes up often.
", "time": "2025-09-30T09:54:33.000Z"}, {"author": "Suresh Krishnan", "text": "We will see you in 66 minutes
", "time": "2025-09-30T09:54:48.000Z"}, {"author": "Farzaneh", "text": "I think we have mentioned that too (what Ekr said). preferences should not be confusing the developers and crawlers, thats why it should be narrow and clear. otherwise it will have unintended consequences. broad preferences can be mentioned through other means ... like terms and conditions etc.
", "time": "2025-09-30T09:54:50.000Z"}, {"author": "Max Gendler", "text": "Farzaneh said:
\n\n\nI think we have mentioned that too (what Ekr said). preferences should not be confusing the developers and crawlers, thats why it should be narrow and clear. otherwise it will have unintended consequences. broad preferences can be mentioned through other means ... like terms and conditions etc.
\n
I certainly would like us to talk about ToS and TaC - but I feel that's an attachments question not a vocab question?
", "time": "2025-09-30T09:56:10.000Z"}, {"author": "Brendan Quinn", "text": "For anyone who missed Mark's announcement, the next session (starting in an hour after the lunch break) will use a different meetecho link: https://meetings.conf.meetecho.com/interim/?group=f5068eaa-7ce2-404b-864c-5a872d1b5ba1 see you over there!
", "time": "2025-09-30T09:56:45.000Z"}]