Re: [Gendispatch] New Version Notification for draft-thomson-gendispatch-rfc-derivatives-00.txt

[Martin Thomson <[e-mail]>]

On Fri, Sep 29, 2023, at 05:52, Brian E Carpenter wrote:
> On 29-Sep-23 05:29, Joel Halpern wrote:
> ...
>> History suggests that we have already solved this problem when there is a clear resolution.  
>
> Exactly. I don't see why we would want a generic policy. There's a real 
> and present danger that such a generic policy would lead third parties 
> to believe that they can produce their own pet alternative protocols 
> *using the same name and IANA assignments* but that don't interoperate.

We do that today by using copyright as a tool to force the Trust into a position as gatekeeper.  As we've established, this only rarely works (because you don't have to copy text to fork IETF work) and it puts the Trust into a position of acting as gatekeeper.  I understand that the right thing tends to happen, but that's not something that is a property of the system, only an accident of having the right people in place at the crucial moment.

My experience with IANA registries suggests that this sort of gatekeeping function tends to produce worse long term interoperability than a looser policy.  The history of RFC 8447 is one where we had to adopt more permissive policies because people were deliberately avoiding the IETF.

The same applies here: the bludgeon we have is not sufficient to stop people from doing the things we would rather not do, but it is likely stopping people from doing things that we would be happy to do (i.e., what the exceptions in RFC 6716 allow; the Debian policies).

A fork of a protocol that uses the same codepoints and different protocol definition (i.e., the RFC 5704 problem) has a defense, which is the same one that was used at the time.  The best defense against that is not the application of copyright law, but the reputation and good name of the IETF.

> As the draft says of the current rules: "This could unduly give the 
> IETF an monopoly over the maintenance of protocols that are published 
> as IETF documents." It isn't "unduly". It is the absolute intention of 
> the "no derivative works" policy. 

Yes, that was clearly the original intent, but the point of that text is to argue against that intent.