[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Fwd: [ppml] Policy Proposal: IPv6 Assignment Size Reduction

Brian E Carpenter wrote:

On 2007-10-31 05:39, Stephen Sprunk wrote:
Thus spake "Mark Smith" <ipng@69706e6720323030352d30312d31340a.nosense.org> 
On Mon, 29 Oct 2007 22:31:10 +0100
Iljitsch van Beijnum <iljitsch@muada.com> wrote:

I thought this would be interesting reading for the working group.


I think the best thing to do is ignore it. If it gets any traction in
ARIN, then we might have to do something.
It doesn't appear to have much support at the present time. OTOH, those of us active in the ARIN community would appreciate pointers to RFCs that say this idea is or isn't valid so that we have as much guidance from the IETF as possible. 

I don't think there is anything that black and white. I think
the best discussion of the issues is in
http://tools.ietf.org/id/draft-narten-ipv6-3177bis-48boundary-02.txt
which is unfortunately expired. Maybe its authors would care to
update it?


Thanks for the pointer, that's a well balanced summary.
There are a few salient points in that however, that I think should be emphasized. (I'll try to avoid context issues, while balancing the need to keep thing brief.) 

  [...] A key
  goal, however, is to avoid the need for a site to renumber into a
  smaller number of subnet bits when adding a new prefix. This could be
  achieved, for instance, by having it be easy for an end site to
  obtain an address block of the same size (or larger) as any existing
  assignments it already has.
The main thing is finding some way to provide assurances to end-sites, that their subnetting plan, will survive renumbering. This is something the RIRs could do with policy, IMHO. Recommend to ISPs that they accept customer requests that are consistent with current policy at the RIR, and 
which were approved by previous ISPs.
And, an observation here is, the likelihood of a new ISP giving a site a prefix compatible with their 
existing usage, is *greater* when their existing usage is a *longer* prefix.
This itself is an argument in *favor* of permitting assignments of longer prefixes. 

  [...] Thus, it
  remains highly desirable to provide end sites with enough space (on
  both initial and subsequent assignments) to last several years.
  Fortunately, this goal can be achieved in a number of ways and does
  not require that all end sites receive the same default size
  assignment.
And my observation is - no one knows better than the end site what its needs for several years will be. Having the end site suggest an allocation, and how it will be used, is the most sensible approach.
Making a larger number of choices available for the end site to select from, encourages thoughtful 
approaches to networking plan.


IMNSHO Brian Dickson's case doesn't hold water. If ISPs are using
careless allocation techniques and thereby wasting space, I don't think
that should be fixed by depriving IPv6 users of their normal /48 or
possibly /56 assignments for small sites. We didn't expand the address
space to allow ISPs to become sloppy at the expense of users. We expanded
it to relieve pressure on both ISPs and users.

Let's not forget basics. /48 is 65536 times bigger than the raw IPv4
address space. Just how *bad* at allocation do we have to be to run
out of /48s???
I've never argued that end sites are likely to return to ISPs for address space.
I have, however, argued that ISPs returning to RIRs for PA blocks, is bad for the Internet. 

And bad for *other* ISPs.
It's like "Other People's Money" - easier to spend than your own. When banks go under, 
it's the bank's customers, not their shareholders, who tend to suffer most.
When ISPs use up router slots, it's mostly *other* ISPs they hurt, i.e. their competitors. 

The ones who have the most to lose, are the ones most likely to behave well.
(It is similar to the nuclear doctrine of Mutual Assured Destruction. Nobody wins.)
It's the ISPs whose equipment *isn't* in jeopardy, who are more likely to behave irresponsibly. 

In fact, behaving badly is actually rewarded, by reducing competition.
Avoiding repeat PA assignments to ISPs, is the *primary* motivation for the proposals, both at the IETF (6man) and ARIN (ppml). 

Brian