Location Information Server (LIS) Discovery Using IP Addresses and Reverse DNS
draft-ietf-geopriv-res-gw-lis-discovery-08

1) This document rightly states that the approach documented is brittle and could possibly do more harm than good.  The right thing to do is use the approach described in RFC 5986.  What I want to discuss is why does section 4.3 say:

   "The DHCP method described in [RFC5986] SHOULD be attempted on all
   local network interfaces before attempting this method."

Why is this not a MUST?  In what situation(s), would a device not perform the RFC 5986 approach first?

2) I think it should be noted that this approach will most likely not work if a device is configured to use a public DNS (e.g., 8.8.8.8).  The question that arises though, is how would the LIS discovery process *know* that the device is configured to use an alternative DNS server?

first support adrian's dicuss and I do belive that it is discuss0-worthy by our current criterion...

Second I deeply concerned (though perhaps I shouldn't be): that what amounts to classful addressing is being proposed for embedding in DNS names. how is that not a harmful precident? I have visions of that being reused in totally inaprropiate fashion, driving address/prefix assingment procedures and so on.

I disagree with the WG's assessment that this is an Informational document. Section 4.2 is protocol. This document specifies a series of queries to determine a LIS based on an IP address of the client. It proposes how to form the query, the types of responses to expect, an algorithm for additional queries based on portions of the IP address, etc. I simply can't see how this is not a Proposed Standard.

aside: I wonder what the average number of discusses for a geopriv draft is.

From the secdir review:

The authors mention that RFC 5986 is the preferred method whenever it is supported.  I believe that this is because it is more accurate.  First question: am I correct in believing this?  Secondly, is there any way an attacker could a) cause
a device to identify the wrong domain name and b) if a device identifies an incorrect domain name (whether or not the
attacker can cause this to happen) is there anyway an attacker can exploit that?

Question 2a) may be difficult to answer because the draft does not mandate any particular solution. Indeed it cannot, because the solution used will depend on what is supported locally.  But if the answers to questions 1 and  2b) are "yes"
it might be worthwhile to point this out in the security considerations section as an additional reason to use the
solution  given in  RFC5389 whenever it is available.

(1) I don't get how its hard to configure a LIS name/addr
but easy to configure a STUN server - can you explain?

(2) Doesn't this require an ISP/access network to likely
be able to answer DNS queries for all sorts of 1918
derived names? That seems undesirable - why is is justified
here?

I agree with Adrian's discuss as I had the same concerns when reading the text.