>I've included the submission announcement below. I realize that this is a >woef >ully late submission for discussion at IETF 95, since I mostly wrote it >yesterda >y, _at_ IETF 95. However, I think it may serve as a better starting point >for >the discussion than the current problem statement draft, so if you have time >to >read it, I would genuinely appreciate it if you could give it a look.
One thing I sort of missed in your draft is a clear distinction between standards track protocols and other protocols. I think that for standards track protocols, the normal IETF consensus process is likely to produce something reasonable. Nothing is perfect, but the examples you gave (.local and .ipv4only.arpa are both quite reasonable uses for the name space). If the IETF feels a need for more non-DNS naming protocols, then probably there will be a discussion in the relevant working group on how to be co-exist with DNS. The problem starts with non-standards track protocols. For just about all protocol parameters (from IP versions to port numbers to MIME types, whatever) the IETF is the only organization that can create a registry. Names (and also addresses and ASNs) are different. There exist organizations outside the IETF that deal with those. In fact, there is quite a bit of history already in some programming languages (for example java) to just register a DNS domain to get a private part of the global name space. So anybody who wants to play with an experimental naming service can just register my-naming-service.net. And use that string in any name switch code. If so desired, additional DNS types can be registered to mark a delegation to the new protocol. This does not require any involvement of the IETF in the registration of the name. Of course, having to maintain a DNS registration is way less convenient than having a entry in some special registry. But why should the IETF be concerned with that. _______________________________________________ DNSOP mailing list [email protected] https://www.ietf.org/mailman/listinfo/dnsop
