draft-ietf-ngtrans-header-trans-02.txt progression

[Harald Tveit Alvestrand]

At 12:21 16.08.98 -0700, Bob Fink wrote:
>ngtrans folk,
>
>Erik Nordmark and I feel that the most effective track for the header
translation draft is Experimental RFC.
>
>I also propose this for other less production ready standards as well to
get them moving along.
>
>Please ruminate on this either to me or the list.
>
Ruminations coming right along.......

there are really 2 kinds of transition strategies:

- Those that can be deployed locally, and the "outer world" sees nothing
  except IPv4 networking and/or IPv6 networking. I think most of the
  NAT-based drafts are like this.
- Those that need some piece of global deployment in order to be effective.
  Dual-stack with AAAA/A record choosing and automatic-tunnel drafts
  require some of this, I think.

For the first kind, having multiple competing proposals is really
a Good Thing. Each organization needing to test something can take its
pick, try it out, and see if it works. It's likely that we will end up
with a small number of proposals that work, but have different scaling
and/or deployment characteristics.
Experimental publication should be encouraged, with standards track
processing once it's proven that it works; merging of drafts isn't
important as long as solutions are clearly intended for distinct parts
of the problem space.

For the second kind, we need to be much more careful; global mistakes
are *very* expensive to recover from - even if we don't factor in the
loss of prestige of the IETF.
Here again experimental publication is a Good Thing, but one needs
much more care in describing how one tells an experiment from a production
network (the IPv6 Testing Address Allocation was one such marker, to pull
in an example from another context), and the eventual target is to have
a sharply limited number of strategies (ideally zero or one) that need
to go standards-track and deployed on the Internet.
After all, we only have one Internet to deploy global-scope solutions on.

My preliminary conclusion: Yes, Experimental publication is a Good Thing.
Drafts need to identify clearly their scope of visibility, and their
scaling properties; once they've done that, and the authors are pretty
sure it works - SHIP IT.

                     Harald A



 
Harald Tveit Alvestrand
IETF Area Director, Operations and Management