At 14:36 +0100 12/27/05, bert hubert wrote:
On Mon, Dec 26, 2005 at 08:05:06PM +0900, Hideshi Enokihara wrote:7.2 Dead / Unreachable Server (OPTIONAL)(...)A server MAY cache a dead server indication. If it does so it MUST NOT be deemed dead for longer than five (5) minutes. The indication MUST be stored against query tuple <query name, type, class, server IP address> unless there was a transport layer indication that the server does not exist, in which case it applies to all queries to that specific IP address.For what it's worth, I consider this section to be meddling. It turns out differing 'dead/unreachable' strategies have differing merits, this should not be set in stone. I missed the standardisation process, but maybe somebody cares. The PowerDNS recursor reserves the right to tune its server selection process regardless of this RFC.
The above is why I use "interoperability" as the measuring stick against which judge the words in an RFC.
The history of documentation in the IETF is thicker than the rules of process. To shed some background on documents like the one above, "once upon a time..."
At a meeting, I forget which but one about 5 years ago, there was a queue of speakers from ISC on the agenda to talk about a bazillion extensions to DNS and descriptions of enhancements that they encoded into BIND. The motivation of these guys was not world domination but to be good citizen engineers and document for all the world to see "best practice." I don't just mean best operational practice, but also the best way to go about implementing something.
I leaned to a neighbor and remarked that this was rather admirable of them - trying hard to fully expose their technology as open source folks - even though they seemed to be bent on monopolizing the agenda. I made a disparaging remark about a competing commercial vendor to contrast this.
What I want to say is that a lot of documents on DNS in the past ten years or so are written from the perspective of (over) eager engineers wanting to make the world a better place. One fault in the work is that they optimized for just the one environment in which they toiled - the open source Internet. Keep this in mind when judging the works.
What failed the IETF then were two things. One is that no alternate view (of what DNS should be) was taken seriously. Note that I did not say there were no alternate views, nor views aired, but that the WG as a whole did not take other views seriously enough. The other failure was in process, quite a few documents got through because engineers value progress over due diligence.
No one has ever "standardized" the resolution process in DNS. I am unconvinced that this is a technical problem. (It could be a policy problem.) The way an process finds an answer to a query is not an interoperability issue, the protocol for doing so is. Ohta's point about burning in timers to application protocols is interesting and right.
DNS uses UDP. UDP is unreliable. UDP packets get lost, as a result the sender has to know when to repeat the transmission. This is not a DNS issue, it's a UDP issue. DNS ought not then specify timers for this, but it is a concern for the datagram management element.
The documents on the shelf are there, I am not issuing a recall request. I just want to point out that you need to read them in the context of the era in which they were written. Future documents editors need to be more careful, as well as reviewers. Standard track documents need to have a standards tone, other engineering ideas ought to be put in non-standards track documents.
I think it's a shame that we sometimes call a document "standards track" because it is updating a document that is standards track. Especially when the original document muddled standards and "what we think is better engineering" ideas.
-- -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- Edward Lewis +1-571-434-5468 NeuStar 3 months to the next trip. I guess it's finally time to settle down and find a grocery store. -- to unsubscribe send a message to namedroppers-request@ops.ietf.org with the word 'unsubscribe' in a single line as the message text body. archive: <http://ops.ietf.org/lists/namedroppers/>