Marco Schmidt wrote on 26/01/2021 14:43:
We don’t think a policy is necessary for this work, but we strongly believe we need an approach similar to NWI-5: "Inform the object holder that these IRR object(s) refer to an unallocated space. We will then delete them after a reasonable notification." This gives people time to react in cases of incorrect de-registration. We will initiate a discussion with the other RIRs to find a joint approach on how we deal with similar incidents in the future.
sounds good.
We have also looked into publishing a list of deregistered resources and this will require some review internally if we plan to go down that route. A more straightforward approach might be for the RIRs to proactively highlight where we see objects in each others IRRs.
Address space can be reallocated after deregistration, so unless the various db objects in the IRRDBs can be deleted at ~the time of deregistration, then there's the possibility that stale IRRDB objects would persist, and it would be difficult to retrospectively figure out whether or not they should be there. Also, there are other IRRDBs out there besides those run by the RIRs. If all the RIRs were to create a deregistration timeline, then this would make it possible for any IRRDB operator to work out canonically whether an resource object is invalid due to previous deregistration. Nick