Rationale problems

[JFC Morfin]

At 08:10 06/12/2008, Harald Tveit Alvestrand wrote:
>If this isn't a concrete example of a problem, I don't know what is.

It certainly is! moreover that due to the hysteresis of the network 
being counted in years rather than in days like for propagation. So, 
it means that several releases may simultaneously be at various 
degrees of deployment, increasing the problems and the list of 
possible conflicts.

This could also be considered together with another interference by a 
new parameter with the DNS: time. The same as IDNA is a new network 
vision, DNSSEC introduces absolute time as a reason why a domain name 
may work at a place and not at another one.. These two new parameters 
can each depend on several authorities (root, zone, local). I fear 
that these kinds of problem do not add but multiply both troubles 
possibility and troubleshouting complexity.

May I suggest we try to keep things simple in investigating and 
thinking about them a little more even if it takes time (everyone 
knows I am impatient about this work to be completed). If 97.5% of 
the requests to the root servers are illegitimate, and some for 20 
years, it means that the real life DNS deployment and management is 
not that simple. Actually many suspect DNS is a mathematical chaos 
situation. Adding any small complexity is always a big risk. Already 
ICANN wants to add thousands TLDs.

I must say that I was not appealed by IDNA as not bringing all the 
answers I think users expect, but now I am quite unconfortable due to 
the adding complexity from areas outside of IETF competence and 
control. IETF is specializing in computer to computer protocols. 
Unicode is specializing in eye computer assisted presentation. 
Languages are brain to brain protocols.

I started having a very quick look at the new documents distribution 
as I promised. I was in favor of extracting rationales from protocol 
descriptions. I start being afraid the mix protected us and we might 
have now a split between what we should do and what we are 
doing/patching. Has someone given a try to a modeling of IDNA 
permitting to "see" how it works?

I am certainly not opposed to a systemic approach, but the DNS is an 
analytic hierarchical architecture. I _may_be_wrong_ and confused by 
the "work in progress" situation, but I feel as if we were just 
seating in between the systemic and the analytic chairs  without a 
clear general architecture picture?

jfc