To:
dnsop@cafax.se
Cc:
edlewis@arin.net
From:
Edward Lewis <edlewis@arin.net>
Date:
Thu, 4 Jul 2002 23:07:04 -0400
Sender:
owner-dnsop@cafax.se
Subject:
support for draft-ietf-dnsop-serverid-00.txt
I'm doing some work on lame delegations and am beginning to think that there is a strong need to require/recommend that a server respond to a liveness query. The draft http://www.ietf.org/internet-drafts/draft-ietf-dnsop-serverid-00.txt suggests something like this. It calls for a way to identify which server is answering a question. My need is a bit more simple. I just want to know if the IP address I am testing has a DNS server (at port 53). This need is derived from the last question I posted to the list - whether a server is allowed to answer to just a few zones and drop all other queries. The reason this is a problem is - I am testing servers for lameness to identify what is the best way to limit the impact of lameness on applications (that don't handle lame answers well). (In other words, I'm not doing this for my health.) In my first runs I would look at answers from servers, and if no answer cam back, I would mark the server down. This worked until I ran across an implementation selectively answers to queries. Now my code is testing all zones, even if the servers are "down." The result is that the test takes a lot more time. So, what I would like to see is a simple liveness query - not even a "version" or "identity" of the server, but just a query/response that behaves like ping. (With attention to the broadcast ping problems.) This would allow testing to happen faster, and with less impact on the network. I see this as bolstering the ability to troubleshoot DNS, i.e., making it more resilient. I would hope that this addition would be acceptable to even the most minimalist implementer out there. -- -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- Edward Lewis +1-703-227-9854 ARIN Research Engineer