[ai-control] Re: -04 drafts for evaluation

[Mark Nottingham <[e-mail]>]

Just to add some context to what Suresh has said -

I think everyone is aware of the concerns that some have about keeping the 'top-level' category. Concerns have also been expressed by others about removing it, so we still need to discuss the nuances of those views so that we can try to come to consensus on an approach. Neither objecting nor holding a vote (or straw poll) is sufficient to resolve the issue: issues in the IETF are resolved through discourse and understanding.

In Zürich, it was felt that we might have a better chance of making progress on the top-level category if we could refine the other categories (what are now in the draft as foundation model training, ai-output, and search) and come to some agreement about their semantics -- the thinking being that doing so might help people become more comfortable with removing it. Alternatively, we might be able to refine the semantics of the top-level category in a way that makes it less of a concern to those who want to remove it. 

That's why we've been focusing on the ai-output and related terms, and why we haven't yet made progress on the top-level category. Rest assured we will get to it. In Montreal, we hope to get a sense of how the changes so far have been received, so that we can assess how solid a basis they are for further work (including the status of the top-level category).

Cheers,


> On 31 Oct 2025, at 12:20 pm, Suresh Krishnan <[e-mail]> wrote:
> 
> Hi Jo,
> 
>> On Oct 31, 2025, at 12:03 AM, Jo Levy <[e-mail]> wrote:
>> 
>> Suresh, 
>> A 1 minute poll at an in-person meeting  which not everyone could attend should not be used to assess consensus. My understanding is that in person meetings aren’t part of the mailing list, which is the vehicle used to discuss and reach consensus. 
>> 
>> Also, as your post indicates, very few people voted in that poll. As I recall, for the one I was able to vote it, the poll was open for less 30 seconds.
> 
> A poll is specifically not a vote. A poll in meetecho is simply a hybrid/remote friendly version of a hum. Please do take a look at RFC7282 (specifically section 4. for some background on humming). To quote from it
> 
> “ Quite often, a chair is faced with a room full of people who seem to be
>   diametrically opposed on some choice facing the group.  In order to
>   find a starting place for the conversation, it can be useful for the
>   chair to ask for a hum to see if one of the choices already has a
>   stronger base of support than the other (or any significant base of
>   support at all, for that matter).”
> 
> 
>> 
>> I also don’t see the names or identities of those who voted. Was it anonymous? Assessing consensus on an anonymous poll seems inconsistent with  IETF policies.
> 
> The “hums” as in the polls *are intended* to be anonymous (we as chairs do not see the identities either). Again look at the reference I quoted above for some context (mainly that this is *not a vote*). The numbers do not matter, the fact that there are people objecting does.
> 
>> 
>> Moreover, there was no consensus to change the original top level category from Text and Data Mining to “Automated Processing.” In fact,  I don’t believe the term “automated processing” was even discussed in this mailing list until after  it appeared in a new draft. 
>> 
>> The ability of the authors to insert new  provisions without any discussion or consensus amongst the mailing list, and then require consensus to remove it, turns the consensus model on its head. 
>> 
>> 
>> 
>> The original logic, as expressed by Paul Keller, was to create a carve out of Text and Data Mining for AI related uses—not to create a signal for blocking all of text and data mining and crawling, and not to recreate robots.txt.
>> 
>> Finally, there was no discussion, let alone consensus, of creating a preference signal for “bots”, once again untethered to AI in any way. 
>> 
>> Frankly, this type of selective reliance on the concept of “consensus” calls into question the integrity of this process.
> 
> Huh, what? Who said there was consensus on this text. I specifically wrote "As far as we can see we don’t see clear consensus either way, and this is something we need to resolve before the drafts can progress from the working group.”. Are you saying that you prefer the TDM definition from the -01 draft instead? 
> 
> Thanks
> Suresh
> 
> -- 
> ai-control mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]

--
Mark Nottingham   https://www.mnot.net/