On Jan 30, 2020, at 5:50 PM, Paul Vixie <paul@redbarn.org> wrote:
On Thursday, 30 January 2020 13:56:04 UTC Karl Reuss wrote:
On 1/29/20 8:33 PM, Paul Vixie wrote:
we have dispensed with 'site' and 'location' as relationships which would define a 'portion of' as a 'root server instance', and that's good (says me). Personally, I'm not ready to give up on location/site as part of the RSSAC definition for instance. When we're at an RSSAC meeting discussing instances, location is the primary way we identify them. I really liked the plain language Wes used in his recent definitions and i'm concerned your more technically accurate definition will confuse much of our RSSAC audience.
If folks really like "site", we could go back to something close to Wes' original wording ("...that serves root data from a particular site...") with: A root server instance is the portion of a root server operator's infrastructure that serves root data at one site using one of the IP addresses associated with a root server identifier. The reason I didn't do that in the text I proposed is because we would probably have to define "site" because...
as before, there have been, are now, and will be cases where two instances of the same identifier are in adjacent racks in some data center that has more than one internet exchange present. location will be the same. SFO1 and SFO2 might be in different parts of the san francisco airport's region, or they might be in the same rack but with disjoint upstream connectivity.
Now that RSOs are partnering with cloud providers to greatly increase the number of instances, it is quite possible that an RSO would have two instances in the same datacenter, but with very different routing being used for each.
this is a topologic location difference without a geographic location difference. it is the topologic location that matters.
Is it important to the reader of this document? Sure, it is important to someone debugging queries getting unexpected answers, but that is not the target reader here. I remain unconvinced that we need to tie the general definition to a topologic location. For some discussions, we need to, but in general all that matters is that a request got to an instance based on an IP address associated with the RSO. --Paul Hoffman