Jinmei,
Thanks for your effort of revising the document. I've reviewed the latest draft, and have some comments. (I must confess in advance that I don't fully remember the past discussions on this document, and that I may be repeating points already discussed before.)
I have a few thoughts on your comments based on my memory of the history :-) I was partially responsible for the previous version.
Bob
Regarding discrepancy with our (KAME/BSD) implementation, whereas I've not done "line-by-line" checks between the spec and the implementation, it seems the implementation supports all the features described in the latest draft type/code/Qtype-wise.
However, - our implementation currently does NOT delay the response to an anycasted or multicasted query.
The question here is the delay useful. To me this would seem useful for multicast queries, but I don't see the need for ones for anycast.
- it does NOT support the proxy responder service. - it still supports the "Supported Qtypes" Qtype which is now marked as "unused" (as far as I know this is just because we've not fully followed recent changes of the spec).
Other general comments follow:
1. In section 2, this document is too specific about particular implementations:
[...] An example of a IPv6 debugging tool using IPv6 Node Information Queries is the ping6 program in the KAME, USAGI, and other IPv6 implementations <http://www.kame.net>.
I'm afraid this is not very appropriate for a general specification document. Instead, we might say:
This was there purposely. At one point there was a lot of push back on moving this forward and it was useful to be very specific about where this was implemented and being used.
[...] An example of an IPv6 debugging tool using IPv6 Node Information Queries is the ping6 program in some platforms.
2. it seems to me that the notion of "proxy" is suddenly introduced in Section 5. I think it should be clearly defined beforehand, e.g., in Section 3.
3. In Section 3, the draft reads:
The Query concerns a "Subject Address" (which may differ from the Queried Address) or a "Subject Name".
We may also want to note that the Subject Address is not even an IPv6 address (i.e., it can be an IPv4 address).
4. I'm afraid the reference to the MD5 spec (RFC1321) in Section 5 should be a normative reference in this context. (If icmp-name-lookups is expected to be published as a PS, this may cause a problem since RFC1321 is informational.)
I don't think that would be a problem.
5. I don't see why we need the TTL field for Node Name, Node Addresses, and IPv4 Addresses if it MUST be always 0. Perhaps this is just for backward compatibility to existing implementations, though...(if so, I believe we should explicitly note that, since otherwise future readers will have the same question)
It is left for backward compatibility. I agree it would be good to explain this.
6. In section 5, the draft reads:
[...] If the Queried Address was anycast or multicast, the source address for the Reply SHOULD be one belonging to the interface on which the Query was received.
I don't see the strong reason in this case for choosing the source address on the receiving interface. If we need to give guidance about the source address in these cases, I think we can (or even should) just refer to RFC3484.
7. Should we still need the "S" and "C" flags for the Node Addresses Query even though site-local unicast addresses and IPv4-compatible IPv6 addresses are now deprecated?
I think it is useful to leave them for backwards comparability. We might want to add some text explaining this.
8. RFC2535 (DNSSEC) was updated by RFC4033-4035. We may want to update the reference as well.
-------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
