-> New RFCs re incremental DNS update by John R Levine -> Re: New RFCs re incremental DNS update by Matthew James Marnell -> Re: New RFCs re incremental DNS update by Kent Crispin ---------------------------------------------------------------------- Date: 29 Aug 1996 18:03:28 -0700 From: "Richard J. Sexton" Subject: Re: Lower The Price Of Admission >2. Guardian and the new InterNic WWWeb pages are cool. >I just wish InterNic would make the Guardian stuff public domain, >for the benefit of registries of all levels... If they're going public, they's zero chance of that happening. We'll have to re-write it. Not a big deal. ---------------------------------------------------------------------- Date: 29 Aug 1996 19:08:57 -0700 From: John R Levine Subject: New RFCs re incremental DNS update I just noticed RFCs 1995 and 1996, dated yesterday, which are about incremental update of domain servers. The model is that the master server sends its slaves a NOTIFY message when something's changed, then the slaves at their convenience send back an IXFR request to ask the master for the new records. IXFR is like the existing AXFR which transfers an entire domain, but just transfers the changes. This looks like a fine plan to reduce the amount of domain transfer traffic, but not of much help to make the DNS a shared-update database, since it looks (upon a fairly quick reading) like each slave is still statically configured to know what masters are authoritative for what. Regards, John Levine, johnl@iecc.com, Trumansburg NY Primary Perpetrator of "The Internet for Dummies" and Information Superhighwayman wanna-be ---------------------------------------------------------------------- Date: 29 Aug 1996 21:54:31 -0700 From: Matthew James Marnell Subject: Re: New RFCs re incremental DNS update :>I just noticed RFCs 1995 and 1996, dated yesterday, which are about :>incremental update of domain servers. The model is that the master :>server sends its slaves a NOTIFY message when something's changed, then :>the slaves at their convenience send back an IXFR request to ask the :>master for the new records. IXFR is like the existing AXFR which :>transfers an entire domain, but just transfers the changes. Can someone tell me what the actual statis of the IANA "announcement" that came out a little while ago is? Also, if you hadn't been watching, or you didn't get the spam from NUMmaster, there is a new I-Draft regarding the .NUM registry, available at your local I-D repository. I'm just waiting to see if the IANA "announcement" sneaks by as a RFC. Matt ---------------------------------------------------------------------- Date: 29 Aug 1996 21:54:52 -0700 From: Kent Crispin Subject: Re: New RFCs re incremental DNS update John R Levine allegedly said: > > I just noticed RFCs 1995 and 1996, dated yesterday, which are about > incremental update of domain servers. The model is that the master > server sends its slaves a NOTIFY message when something's changed, then > the slaves at their convenience send back an IXFR request to ask the > master for the new records. There are multiple masters -- each registry is one. The "central" root dns server is a slave to all the registries, and is configured to respond instantly to a NOTIFY. > IXFR is like the existing AXFR which > transfers an entire domain, but just transfers the changes. In this particular case the changes are very small. The registry "master" servers are really only stub servers -- they don't need the entire database, and they don't need to support the entire DNS protocol. And they don't have to keep track of any changes locally, because by definition of their function everything they send is a change. Vixie says the implementation will have no trouble supporting multiple transactions per second, which is plenty high enough for our purposes. > This looks like a fine plan to reduce the amount of domain transfer > traffic, but not of much help to make the DNS a shared-update database, > since it looks (upon a fairly quick reading) like each slave is still > statically configured to know what masters are authoritative for what. There is only one slave. It is statically configured to know about the registry servers. It also has to authenticate the requests. Remember that the protocol between the registries and the root server for the domain doesn't have to be the same as between all other name servers -- they are, after all, in a special relationship. In the proposal I sent around I said that DNSSEC, IXFR and the others were very *close* to what we need. Putting a different twist on the meaning of "master" and "slave" isn't much of a change to the protocol. Allow me to hand you this convenient bag they keep in the pouch below the tray table, behind the seat in front of us :-). - -- Kent Crispin "No reason to get excited", kent@songbird.com,kc@llnl.gov the thief he kindly spoke... PGP fingerprint: B6 04 CC 30 9E DE CD FE 6A 04 90 BB 26 77 4A 5E