Re: endpoint discovery from reverse DNS [Re: other comments on dr aft-nielsen-v6ops-3GPP-zeroconf-goals-00. txt

[Alain Durand <Alain.Durand@sun.com>]

Pekka Savola wrote:

> But how many of these cases are such that a DNS search-path based 
> approach would not be suitable (due to the requirement to have more 
> control on which tunnel endpoints each are selected by which node, as 
> pointed out by Alain)?

Any large network that spans different locations and uses potentially 
multiple domains.
This is true for large enterprise networks, but also at home if the user 
decided to have its
own local domain advertized through DHCP. For example, at home, my local 
DHCP
server is configured to send a search path as "sun.com" and 
"mylocaldomain.example.com",
but not "myISP.example.com"

So if i was to use the DNS search path in the forward tree, I would 
end-up looking for
Tunnel-end-point.sun.com or Tunnel-end-point.mylocaldomain.example.com 
but never
for the right thing, that is Tunnel-end-point.myISP.example.com.

To paraphase what Rob Austien once said about automatic completion using 
domain search list,
when you do not know what the question you ask is, don't be surprised if 
you don't find the answer.

In other words, you should never believe that is advertized as your 
domain name
is relevant to figure out where you are physically on the network.

   - Alain.