Re: [Crisp] staring points

Leslie Daigle wrote:
> 
> Nice try; wrong tld... ;-)  The IETF decamped from  .int, and
> we now use .arpa for such needs.
> 
> So, yes, SNAPTR could be used (like in DREG); areg-iris.arpa
> could include appropriate NAPTR records to hand off to the
> appropriate server (e.g., iris.nro.net).  As part of its
> caretaking of the standard, the IETF/IAB would undertake
> to make sure IANA kept its DNS up to date for areg-iris.arpa.
> (It's not like it's going to be highly volatile!).
> 
> That speaks to the "enshrined location of the top AREG server"
> issue; it solves some of the issues with it.

While I understand that this is probably politically easier to do, we're 
really talking about moving the "enshrined location of the top AREG server" 
into the DNS. Saying, "start at areg-iris.arpa" isn't really that different 
from saying "start at iris.nro.net", except that the IAB runs it and not the 
NRO.

Which is not to say that I'm opposed to this, I just think we shouldn't kid 
ourselves about what the solution really adds.

> There still seems to be a question about whether there should
> be a unique "top" to AREG, or whether that should be configurable
> in the client.  That doesn't preclude the possibility of having
> some conventionally known good/best starting places,  such as
> iris.nro.net.
 >
> It's the latter that I recall Ted calling for during the
> meeting (and seems to be captured in the minutes).  However,
> the current AREG URI document doesn't seem to make the
> distinction -- it suggests (practically) that there is one
> top, and that it is located at iris.nro.net.
> 
> For myself, I think either the document should be clearer
> that is *one* possible starting point and clients should be
> configurable to allow others, or it should suggest
> how to find the current "top" at any given time (other
> than enshrining a host name).

I don't think anybody thinks that clients should be prevented from having 
their initial query location configured on their side. Clients could even 
have a set of starting points depending on what AREG information they are 
looking for.

But... one of the reasons that the RIRs are so interested in IRIS is because 
clients have to configure these starting points today!

The main reason to want a unique "top" to AREG is that it is actually how 
number management works today - the same as DREG. Clients that follow the 
hierarchy from the top will get the "best" information possible, at least 
from a global view.

I can envision any number of scenerios where a top-down approach is not 
complete. For example, at a university it might make more sense to try a 
local AREG query first. Likewise at an exchange point each ISP peering there 
may provide an AREG server with detailed information to their peers. But I 
am fairly sure these are the exception, and nothing prevents clients from 
taking advantage of knowledge about the local network environment.

I hope the latest draft that Engin submitted over the weekend is clear 
enough to meet the desire for flexibility and having a well-defined default.

-- 
Shane Kerr
RIPE NCC

_______________________________________________
Crisp mailing list
Crisp@ietf.org
https://www1.ietf.org/mailman/listinfo/crisp