Re: [DNA] No FastRA unless Landmark? (draft-pentland-dna-protocol-00)

[Greg Daley <greg.daley@eng.monash.edu.au>]

Hi Sathya,

Sathya Narayanan wrote:
> Greg -
[cut]
> No. 
> 
> Looks like there are two reasons why a host will not include a landmark option
 > (I thought about only the first):
> 1) Host is a non-DNA host
> 2) Host doesn't want to divulge their landmark
> 
> The text in the draft is targetted at the first case, i.e. if the RS message
 > doesn't have a landmark option it is a non-DNA host. The only way to 
help out
 > the host is by sending a CompleteRA because the router doesn't not 
know what
 > question to answer. For a non-DNA host, a completeRA will be only 
partially helpful,
 > if it hasn't moved link and sees its current prefix in the completeRA 
in a PIO -
 > because it can neither recognize the completeRA bit nor the DNAO. 
Sending too-many
 > unicast CompleteRA with the hope that it is useful for the host is 
not good for
 > link-utilization and hence the recommendation.

The question wasn't about completeRA, it was about FastRA.

Non-DNA hosts aren't going to benefit from the RA's DNA content,
but they are going to benefit from a fast response.

So there really shouldn't be a problem with supplying fast
response so long as the solicitation is compliant with
the algorithm for delivering an RA quickly (has a source address,
and can create, or already has an NCE).

(Now visiting the CompleteRA without Landmark solicitation issue).
> Now, I think, the same logic applies for the second case too, even though a DNA-host
 > could receive better benefit from the CompleteRA. The router has 
no-way of knowing
 > whether the sender is a dna-host or not.

This is the essential point for completeRA without landmarks.

Why doesn't it know? because without the landmark you can't tell that
it's DNA capable.

Fortunately, there are plenty of bits left in the RS header (32)
which could be used to identify a DNA solicitation without divulging
the last attached prefix of the host in a landmark option.

Please be aware that I personally consider location privacy to be a
'big issue'.

Greg