Re: [stir] current draft charter

[Dave Crocker <[e-mail]>]

On 6/12/2013 10:23 AM, Henning Schulzrinne wrote:
> The simple semantics sounds similar to the RFC 4744 model for SIP URIs. As discussed in various drafts, this model unfortunately doesn't work well for the domain-less case, such as telephone numbers.


Not definitionally, no, but of course we know it can be encoded as one. 
  Ignoring the politics of that moves the discussion to a comparison of 
the choices among alternatives.

At a minimum, any solution here has to provide cert semantics, encoding 
and operation, query protocol and integrated server operation.  (There 
are probably some additional requirements, but these three will do for a 
start.)

My understanding of the requirements for cert semantics here is that 
it's pretty trivial.  Administering and enforcing it might not be, but 
the basic semantic quite simple.

New protocols are expensive and risky, even when built on top of a solid 
technical infrastructure.

We know that establishing a globally integrated set of server operations 
is risky to the point of being nearly impossible, based on a long 
history of many attempts and very, very few successes.

      (The word "integrated" is the challenge here -- the service being 
proposed here is for a unified number space; servers handling queries 
therefore need to compose into a unified service; for example this is 
quite different from fully independent web servers that might have 
pointers to each other.)

My suggestion is that any proposed solution be considered carefully for 
comparative effort and risk.  The challenge of course is to be fair to 
whatever choice isn't one's favorite.  Perhaps a cliche, but on the 
average we think the challenges for the choice we prefer are less than 
they actually are, and for non-favorites higher than they actually are.

d/
-- 
Dave Crocker
Brandenburg InternetWorking
bbiw.net