At Fri, 3 May 2002 09:49:45 -0700, Steve Deering wrote: > > It's not the routing system "deciding" the choice of servers, it's the > human who causes the host routes to be injected into the routing system. > Instead of a human putting three DNS server addresses into a DHCP > database, a human configures the three DNS servers themselves to > advertise the (three different) host routes. The client still gets to > choose among them, using whatever criteria it wishes.
Only if the binding between the well-known unicast address and the particular name server at that address is stable. If the binding is a dynamic thing, then choices that the client tries to make based on previous history of the server at that address become invalid when the binding changes, and the client has no way of knowing when the binding has changed. > A couple questions, from one who is expert in neither DNS nor DHCP: > Do the bulk of DHCP servers today provide more than one DNS server > address to each client? If so, do "consumer-level" IP devices (PCs, > laptops, PDAs, cell phones, etc.) really make sophisticated choices > among those multiple DNS servers, or do they just pick one and then, > only if that one fails to respond, try another? Depends on the implementation, but all DNS clients are required to be capable of being configured to know about more than one server at a time. Some implementations just round-robin among the servers they know about, some perform timing measurements and try the server that had the best response time in the past for the next query, and so forth. -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
