Re: I-DAction:draft-ietf-kitten-kerberos-iana-registries-00.txt
Jeffrey Hutzelman <[email protected]> Sun, 28 Oct 2012 00:47:01 -0400
| Newsgroups | gmane.ietf.krb-wg |
|---|---|
| Message-ID | <[email protected]> |
On Sat, 2012-10-27 at 12:07 +0100, t.p. wrote: > > Message types are probably not suitable for registration. The sort of > > change that would result in a new message type number would be a > > significant change to the protocol, well beyond what reasonable people > > might agree is a new protocol parameter value. > > > > Message types are identical to the ASN.1 application tag numbers of > > protocol messages that contain them. In any given type of protocol > > message, the msg-type value will never change. New message type > > numbers will only come into existence with new protocol messages, and > > thus be subject to standards action. > > Um; that logic seems to say that anything that is subject to Standards > Action need not be in the IANA registry, which seems to me to be quite > wrong. It would, however, enable us to eliminate much of the current > web site. > > The IANA registry is a major source of information, a centralised > database of anything, in the shape of a name or a number, that you need > to know in order to understand or deal with the Internet. Thus SNMP > puts Security Models and Message Models there, a new one of which would > take a new Working Group and years of work to produce. I am sure that > there are many more such cases in our core protocols. Having a central > source, with references, makes the information usable to everyone. > > So, I think that Kerberos Message Types would make a most useful > registry. No, Tom's point here is that the message type numbers are in fact in exactly the same situation as the protocol version -- in any place where such a number appears, there is only one value that can possibly appear there; any other value is a protocol error, not an opportunity for extension. In particular, message types are not used to distinguish between messages -- they cannot be, since they are contained within each PDU and such distinction must be done before a PDU can be parsed. IANA registries are not intended as "a centralised database of anything [...] you need to know in order to understand or deal with the Internet". They are intended to serve as a mechanism for the management of namespace, to insure that the same name (number) is not reused for more than one purpose. Where there is nothing to manage, there is no need for a registry. -- Jeff _______________________________________________ ietf-krb-wg mailing list [email protected] https://lists.anl.gov/mailman/listinfo/ietf-krb-wg