Re: 6MAN WG Adoption call : draft-jeong-6man-rdnss-rfc6106-bis

[神明達哉 <[e-mail]>]

At Mon, 5 Oct 2015 19:30:47 +0900,
"Mr. Jaehoon Paul Jeong" <[e-mail]> wrote:

> Here is the revised draft reflecting the discussion of link-local addresses
> for recursive DNS servers:
> https://tools.ietf.org/html/draft-ietf-6man-rdnss-rfc6106bis-03
>
> The major changes are as follows:
>
> 5.1.  Recursive DNS Server Option
> ...
> Note:  The addresses for recursive DNS servers in the RDNSS option
>       MAY be link-local addresses.  Such link-local addresses SHOULD be
>       registered into the resolver repository along with the
>       corresponding interfaces (or zones) that received the RDNSS
>       option(s) for them.  The interface names (or zone indices) SHOULD
>       be represented in the textual format for scoped addresses in
>       described in [RFC4007].  When a resolver sends a DNS query message
>       to an RDNSS with a link-local address, it MUST use the
>       corresponding interface.

I guess you're trying to address the following comment of mine:

>> I don't have a strong opinion on whether to allow link-local
>> addresses per se, but if we allow that, I think we should clarify
>> which link (or link zone, in RFC4007 terminology) that link-local
>> address belongs to.  In the case of router advertisements, I guess it
>> makes most sense if it's the link over which the RA is sent out and
>> received.

My impression of that discussion was that for others this is too
trivial to explicitly note.  So I would have been okay with not saying
anything about it, but now you're actually trying to address that
(which I appreciate), I'd like to offer some changes based on the
following observations:

- the above text seems to indicate the common confusion on the
  difference between interfaces and links. For example, technically,
  there's no "corresponding interface" or a link-local address;
  there's only "corresponding link", and you can identify the link if
  you know an interface that belongs to the link as a context, but
  that's not the "corresponding interface of a link-local address".
- using the interface name to represent a link index is an
  implementation specific hack with the implicit assumption of
  one-to-one mapping between links and interfaces (that's often the
  case but architecturally not entirely correct).  so I suspect it's
  not really appropriate for normative and generic text like this

So I suggest revising the above note as follows:

 Note:  The addresses for recursive DNS servers in the RDNSS option
       MAY be link-local addresses.  Such link-local addresses SHOULD be
       registered into the resolver repository along with the
       corresponding link indices of the links that receive the RDNSS
       option(s) for them.  The link zone indices SHOULD
       be represented in the textual format for scoped addresses as
       described in [RFC4007].  When a resolver sends a DNS query message
       to an RDNSS with a link-local address, it MUST use the
       corresponding link.

Or, as we saw in the previous discussion, if many others don't see
the need for a note like this in the first place, I wouldn't be
opposed to removing it completely.

--
JINMEI, Tatuya