> On 19 Jul 2020, at 6:22 am, John C Klensin <john-ietf@xxxxxxx> wrote: > > Extrapolating from the above, it sounds like an interesting > middle ground might be to allow very lightweight provisional > registrations but start a clock on them such that, if a > specification did not appear within a set period of time, they > would be at least marked obsolete and maybe removed from the > registry (the difference being whether the name would be > considered available for reuse). At the risk of more > complexity, perhaps that could be "mark obsolete, allow for some > appeal period that would require proof of active use and at > least a promise of documentation, then remove". One could > further tune something like that to treat I-Ds for a while but > not very long; after that, RFC or some other stable > specification. > > That is just an idea, with lots of details absent and in need of > filling in, but I wonder if something like it would address the > issues you are raising without discarding the idea of quick > registrations to eliminate or reduce naming conflicts. That's somewhat similar to the model we're using for Link Relations and Well-Known URIs (both Specification Required). For them, you can make a provisional registration if there's an I-D or similar ephemeral document, but the expert(s) are empowered to remove it if there isn't progress to a more stable reference. Registrations without _any_ document (however brief) aren't allowed. The experts are also empowered to register (potentially as permanent) values that are found to be in widespread use, after consultation with the community. It's early days, and we may need to add some more process when we start thinking about following through on removing registrations, but so far I like the properties it has. Cheers, -- Mark Nottingham https://www.mnot.net/
