in line...

Stephen Sprunk wrote:
> 
> Thus spake "Brian E Carpenter" <[EMAIL PROTECTED]>
> > Margaret Wasserman wrote:
> > ...
> > > SUBSTANTIVE ISSUES:
> > >
> > > (1) This draft doesn't mention the reverse DNS tree.  Is it expected
> that
> > >      whatever registry assigns these values will also populate the
> reverse
> > >      DNS tree?  Or not?
> >
> > I think it is better to leave this question for a separate document (which
> > certainly needs to be written). I don't think we should delay the present
> > document for this.
> 
> I think this revolves around a basic question of whether information about
> centrally-assigned local prefixes should be available to third parties.  If
> so -- and that is my preference -- then WHOIS information should be provided
> and DNS delegations should be available, though not mandatory.  Optional
> global DNS delegations may be beneficial to those using local addresses to
> communicate privately between organizations because it can remove the need
> for split DNS.
> 
> > Yes. I suggest rewriting the first extract as:
> >
> >  Global IDs should be assigned under the authority of a single allocation
> >  organization because they are pseudo-random and without any structure.
> >  This is easiest to accomplish if there is a single authority for the
> >  assignments.
> 
> This wording still doesn't seem consistent with the possibility of having
> multiple central authorities.  If we are going to allow IANA the discretion
> to assign multiple allocators, then we need to specify how that will be
> handled, i.e. either assigning sub-prefixes to each authority or requiring
> real-time checking between authorities to prevent collisions.

Send text? I meant mine to say that the authority (IANA) is unique - that
doesn't preclude (or require) sub delegation.

> 
> > >      It seems strange to say that there can be no periodic fees when the
> > >      document doesn't mention fees anywhere else...  Perhaps this is a
> > >      leftover from previous versions of the document that did include a
> fee
> > >      structure?
> >
> > No, I think it is a principle and a directive to the IANA that we should
> > keep.
> 
> Agreed.  Perhaps we should emphasize the permanent (i.e. rent-free) nature
> of these allocations earlier in the draft so this statement doesn't seem out
> of place?
> 
> > >      I am not sure that requiring a permanent escrow is consistent with
> the
> > >      idea that there will be no ongoing revenue stream (i.e. periodic
> fees)
> > >      associated with an address.
> >
> > But it's a matter of principle. If the IANA really thinks this is an
> issue,
> > it's for them to say so, not us.
> 
> This is more of a business concern than a technical one, which seems outside
> the IETF's purview.  Maybe under IANA considerations we could add a
> statement about how IANA should require a prospective allocation authority
> present how it can fund indefinite operation from one-time fees?  As has
> been discussed, any competent accountant can figure how to determine a
> one-time fee that equates to indefinitely recurring fees.
> 
> It might be desirable for IANA/ISOC/whoever to set up a permanent trust
> which receives one-time allocation fees and pays recurring fees (from
> interest) to the actual allocation authority; such trust could also be
> responsible for the escrow of registrations.  This would protect against
> bankruptcy of (or subsequent change in) the particular allocation authority.
> I'm not sure this level of detail belongs in the draft, but this is the
> model I had in mind.

That certainly doesn't belong in an IETF document. But saying "no recurring
fee" seems to me to be something we *can* say. And that's the text that went 
through WG last call.

   Brian

> 
> > >      Are you intending that centrally assigned addresses will all be
> > >      generated using the EUI-64 address of a server at the centralized
> > >      registration authority?  What affect would this have on the
> randomness
> > >      of these allocations, and does that matter?
> >
> > Actually the randomness doesn't matter much - we need enough
> > randomness to ensure that people don't imagine these things aggregate,
> > but there is no need for a mathematically "good" randomness.
> 
> It's possible that the generation method could be bottlenecked by the rate
> of timestamp changes if centrally-assigned prefixes become very popular.
> Since the centrally-assigned half of the address space doesn't need the
> level of collision prevention that the locally-assigned half does, could we
> change this from a MUST to a SHOULD for the former?  Any pseudo-random
> algorithm should be sufficient for a central authority, provided it gives no
> potential for aggregation.
> 
> S
> 
> Stephen Sprunk        "Stupid people surround themselves with smart
> CCIE #3723           people.  Smart people surround themselves with
> K5SSS         smart people who disagree with them."  --Aaron Sorkin

-- 
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Brian E Carpenter 
Distinguished Engineer, Internet Standards & Technology, IBM

--------------------------------------------------------------------
IETF IPv6 working group mailing list
[EMAIL PROTECTED]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to