[Ietf-not43] extensibility

[David Blacka]

On Friday 25 July 2003 04:01 pm, Edward Lewis wrote:

> What has raised my hackles is the thought that "reversing the flow" 
> (I mean finding the 'not whois' data for something from the bottom 
> up) is not the same between domain names and IP addresses.
> 
> Given a domain name, since the name is actually a path, I can see 
> it's registration heritage.  I.e., placing SRV records at a specific 
> location relative to the domain name is a pretty reasonable 
> convention for finding where the right authority of registration 
> information might lie.
> 
> This isn't so true with IP address ranges.  A limitation of DNS gives 
> us essentially a choice of /16 and /24 sized zones, with the RFC 2137 
> process to help a bit in un-CIDRizing this, meaning that for an 
> arbitrary range, it's not so clear cut where to put the SRV-like 
> pointers.  (They may need to be replicated 127 times for a /17.) 
> That, and, because there's some question as to the value of the 
> reverse map, esp. in IPv6, there's no guarantee that there is even a 
> place to put the SRV-like things.

Let me try to restate this:  you are saying that because address allocation 
information, unlike domain allocation information, does not need DNS to 
function, and so the reverse tree is broken, clumsy or nonexistent.  Thus, 
using the reverse tree to find CRISP servers is problematic.

Did I get that right?

Assuming I did, and it seems like a reasonable conclusion to me, do you 
have a better suggestion?  do you think that all address searches should 
start from the top (i.e., just don't do SRV lookups)?

-- 
David Blacka    <davidb at verisignlabs.com> 
Sr. Engineer    Verisign Applied Research