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

Re: [dhcwg] DHCPv6 router option



On Fri, Mar 13, 2009 at 01:09:59PM -0400, Ralph Droms wrote:
> David - let me ask for specifics here.  What sections would you drop from 
> the current draft, which would be replaced at a later date by additional 
> RFCs from v6ops?

I don't think I would require the current draft to be changed
materially.  Section 4 contains no normative language, and is
therefore largely informational.  If it were mine, I would have
focused more on an overview of how the information gets from the
server to the client (ORO, placement of the option in the reply
packet (root? IA?)), with an admission at the end of its intended
use.

So my ideal section 4 wouldn't refer to RFC 4861 at all; except that I
understand its current use as a vehicle to explain the fields'
contents and possible use to likely readers (the author's intended
audience).  To those ends, I think the authors should have freedom
here.

And of course, I think you need to move the RFC 4861 reference to
an Informative References section, since I do not believe there has
been any normative language referring to ND currently, and I would
suggest against adding any.


So long as we're talking about me;

I have also been thinking that I would drop the router lifetime
values, preferring instead for the fields contents to be "valid
as long as the client holds its current lease (basically the longest
valid lifetime of any addresses granted)", but I don't know if
this interferes with some intention to support these options via
Information-Request messages, so I have been witholding comment.  I
planned to wait and see the presentation before suggesting
alterations.

I do know that our software, as it operates at userspace and not in
kernel space, does not seem to have reliable mechanisms to store
router lifetimes to the kernel (depends on architecture, on some it
is merely undocumented), and we probably would on some platforms tie
the insertion and removal of any routes to IP address lifetimes
instead until such time as kernel keepers make this configuration
possible.

I would prefer not to have to add extra complexity to track the router
lifetime separately, at any rate.

The idea is that this follows the same way we use or invalidate other
configuration state in the lease - such as nameservers or domain
search.  They are valid until the last address expires, and they are
updated (old-removed, new-inserted) upon every Renew or Rebind.

-- 
David W. Hankins	"If you don't do it right the first time,
Software Engineer		     you'll just have to do it again."
Internet Systems Consortium, Inc.		-- Jack T. Hankins

Attachment: pgpb2zIq5i1DG.pgp
Description: PGP signature