Dave and I appear to be in agreement here. I have spent quite a bit of time thinking what the consequences of a default by a registry of the IANA, ISBN, EUI (aka MAC address) type. My conclusion is: virtually nil. Let us imagine that the folk who assign EUI numbers decide that they are not going to allow Iran or North Korea to register a number. This is far from theoretical, in fact it is US law today. The Iranian manufacturer simply announces that they are going to use a particular prefix and starts making ethernet cards. If you are another manufacturer there is simply no way that you are going to accept the unauthorized Iranian assignment. Similar strategies apply to IANA assignments, including assignment of IPv4 address space. It is best if nobody attempts to employ IANA as an Internet choke point, it is not. The issue is unlikely to come up unless the resource being allocated is finite: Port numbers, DNS RR codes, protocol numbers. My personal view is that we should develop an Internet architecture that allows an infinite number of new protocols to be deployed without consumption of scarce resources, i.e. port numbers of DNS RRs. I think that it is entirely defensible for IANA to be parsimonious with such assignments because they are essentially the fabric of the Internet. Assignment of non-finite identifiers should be a free for all. Anyone should be allowed to apply for a text based label (i.e. algorithm identifier, SRV prefix, ASN.1 OID) at any time with no process whatsoever. Until now we have been dealing with application protocols that almost exclusively deal with the human-machine interaction: Email, Web, FTP, NNTP. Only a handful of protocols are machine-machine. This is changing with Web Services/Mashups/Semantic Web. In the future we are going to see a vast number of machine-machine protocols, only a tiny proportion of which will be standards based. The lack of standards is a good thing, neither the IETF, W3C or OASIS or all three combined is going to have the bandwidth to standardize everything. And unlike human-machine protocols the cost of running multiple machine-machine protocols in parallel is not unacceptable as a general rule. Moreover premature standardization is a bad thing, particularly when nobody really knows what the protocols should be doing. So in summary, the IAB should be charged with identifying the set of finite resources that IANA assigns and propose an Internet architecture in which deployment of new application layer protocols does not cause any of the finite resources to be depleted. > -----Original Message----- > From: Dave Crocker [mailto:dcrocker@xxxxxxxx] > Sent: Tuesday, June 12, 2007 11:49 AM > To: John C Klensin > Cc: ietf@xxxxxxxx; iesg@xxxxxxxx > Subject: Re: IANA registration constraints > > > > John C Klensin wrote: > > Again, there may be exceptions, but I think denial cases should > > require fairly strong (and public) justification. In the general > > case, I believe the Internet is better off if even the most > terrible > > of ideas is well-documents and registered --with > appropriate warnings > > and pointers-- if there is any appreciable risk that it will be > > deployed and seen in the wild. > > > Mostly, I think we (the community) tend to confuse the > coordination role of registration with the approval role of > standardization. > > d/ > > -- > > Dave Crocker > Brandenburg InternetWorking > bbiw.net > > _______________________________________________ > Ietf mailing list > Ietf@xxxxxxxx > https://www1.ietf.org/mailman/listinfo/ietf > _______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf