[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: RFC 5006 status



On Mar 17, 2010, at 09:44 , Durand, Alain wrote:

Now, about this particular document:

  1. The interaction between DHCP & RA is not very well documented.
That could be improved when revising RFC 5006 for proposed standard.  Apple's implementation merges all IPv4 and IPv6 addresses for DNS servers from all the available sources, DHCP4, DHCP6 and RFC 5006.
  1. Why do we need a lifetime in that DNS RA? This create extra complexity for no apparent good reasons
Is the reason for the lifetime field really not apparent?  It was perfectly apparent to me when I implemented it, i.e. so that hosts will stop using DNS servers when their lifetimes expire.  This makes renumbering less painful.  I suppose you could argue that the router lifetime should imply the DNS server lifetime, but that would be a weird departure from the conventional architecture of ICMP6 timeout parameters.
  1. What is the host suppose to do when it gets DNS RAs from **different** interfaces? Should it treat them the same as if they came from the same interface? Or is it that the information is tied to the prefix, not to the interface?
The same thing the host does when it gets DNS server addresses from DHCP servers on different interfaces.  Multihomed hosts have a choice of models for dealing with merging of DNS server configuration: A) merge them all into one list, or B) merge them into a list per interface.  (The RDNSS options are not tied to PIO, so that's not an option.)

In other words, it seem that the general cases beyond a single interface, single prefix are not well documented.

Most of the issues here revolve around configuring multihome/multistack hosts with recursive DNS server addresses from a variety of sources not necessarily RFC 5006, and those issues should probably be addressed separately whether RFC 5006 goes to proposed standard or not.  There really isn't very much here that makes RFC 5006 into a specific problem in and of itself.  We can argue about the lifetime parameter, which I happen to think is a pretty good idea, but others may differ.


--
james woodyatt <jhw@apple.com>
member of technical staff, communications engineering