Re: [DNA] Comments on DNAv6 (draft-dnadt-dna-protocol-00.txt)

[Erik Nordmark <erik.nordmark@sun.com>]

Subba Reddy wrote:
> Hi,
> 
> few comments and questions on the DNAv6 draft.

Thanks for the comments.

> 1. In Section 3.2
> 
> "....    In the case when the landmark prefix is unknown to the
> responding router, the host will receive a "No" answer to its
> landmark question, and also the information it needs to configure
> itself for the new link.  The routers try to include as much
> information as possible in ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ such
> messages, so that the host can be informed of all the prefixes 
> assigned to the new link as soon as possible."
> 
> but as per section 5.1.5, DNAO for learned prefixes must not be added
> in the unicast RAs. i.e, we are not adding as much information as
> possible.

Well, sending DNAOs to hosts when responding yes or no to a landmark 
question doesn't actually provide the host with any (useful) 
information. In order to provide information to the host in a "no" 
landmark RA, the DNAO would have to include the prefix lifetimes and the 
flags.

> 2. While bootstrapping DNA data structures(section 5.1.3), a) will
> the router reply for RSs without landmark option (may be sent by
> other routers) b) will the router reply for RSs that contains
> landmark option (may be sent by DNA host)

I think the current text says that it will not respond at all. This is a 
bit suboptimal when multiple routers boot at the same time.
I suspect we need to figure out how to tweak this.
Suggestions?

> 3. While a router is in multicast sender state (during 3 seconds),
> will it increase the token count or will it consider the token count
> as MaxUnicastRABurst?

The token bucket gets tokens added periodically independently of whether 
its in multicast state or not, as long as the bucket isn't full (already 
at MaxUnicastRABurst)

> 4. In section 5.2.4,
> 
> ".....If the unicast Router Advertisement contains a Landmark option 
> that matches the Landmark Option in the last transmitted Router 
> Solicitation ....."
> 
> Is a case where, a unicast router advertisement comes without
> landmark option is possible? (DNA routers which are bootstrapping may
> just send RA) If possible, some text which tells what to do in such
> cases may need to be added.

We should probably say what to do in any case, since an existing RFC 
2461 router is free to unicast RAs (even if the RFC suggests to always 
multicast them)

> 5. To deal with amplification effect, what about having a threshold
> on number of routers sending fast RA? For example, only first 5
> routers (as per the fast RA rankings) only will reply for RSs (in
> addition to the method specified in the draft)

That might make sense.

I assume the 6+ ranked routers would not be completely quiet, but 
instead schedule a 3 second multicast RA. This is needed in the case 
that the routers annouce different prefixes, and also to make the host 
discover that it has >5 default routers.

    Erik