Mark Andrews wrote: > > Martin Rex writes: > > > > Thanks for mentioning rfc 4074. The stuff in that document matches > > the thoroughly broken behaviour of the IPv6 DNS resolver client of > > Windows 2003 that I had encountered just recently. > > > > IMO, rfc4074 exhibits a significant amount of cluelessness about DNS, > > the "Full Standard" document maturity level, and the realities of > > backwards compatibilities for an incredibly huge installed base. > > So not answering a query because the type is 28 matches RFC 1034 > behaviour? DNS is a query/response protocol. > > So returning NXDOMAIN because the query type is 28 when you have > type 1 in the database matches RFC 1034 behaviour? > > Returning A record content in software written *after* type code > 28 was defined matched RFC 103[45] behaviour? The same servers > shove A rdata into TXT rdata. Everything is a A record.
We started this with the NOTIMP, which is certainly a permissible response for an AAAA query per rfc1035. What I briefly saw the win2003 DNS client doing in a wireshark trace (which I forgot to save before I had to reboot after windows disabled all networking) was visualized as a mixture of "no name" and "server fail" for 23 consecutive AAAA lookups until ~18 seconds later the first A lookup was finally sent. I assume the two locally configured DNS Server were caching(-only) servers, not authoritative ones (at least our current DNS Zones do not contain NS records for any of the two). Besides what you can deduce from the spec itself, you will also have to take into account how that incredibly huge installed base was created. A number of software vendors perform only black-box testing and limited interop testing, so it will not be unusual that your AAAA queries may be the first queries of that kind that a DNS responder might be seeing. So if the behaviour (how to exactly respond to queries for unknown QTYPEs) is neither explicitly specified, nor likely have been part of the usual/common interop tests performed by the vendor, what you're left with might be "ureflected&untested guessing" on part of the implementors to fill those gaps. Bottom line, the receiver must be VERY conservative with assumptions about what exactly can be infered from error responses for situations that are fairly vague in the spec and potentially untested in the installed base. That is called "robustness principle". -Martin _______________________________________________ Ietf mailing list Ietf@ietf.org https://www.ietf.org/mailman/listinfo/ietf