[address-policy-wg] Policy Proposal 2006-02 (IPv6 Address Allocation and Assignment Policy)

Nick Hilliard wrote:
>> The "must be advertised" part causes any non-internet usage to be
>> 'illegal'.
> 
> I disagree.  The text states "This prefix must be advertised within one
> year of the allocation being made".  It doesn't state where the
> announcement should be made.  If the prefix is announced on a private
> network, then the LIR has satisfied their obligation to this clause.

True, one could read it like that indeed. As Randy Bush tends to say
'which Internet'. But that is not the intent of the letter.

If you are reading like that, then the "200 issue" is not an issue
either, as then I will name every mouse in the world a organization and
assign them all a /48.

>> As the number of assignments is removed, there is no way to check up on
>> this, next to there not being any requirement of actually registering
>> these assignments in the RIPE db, thus there is no way to check.
> 
> The alternative at the moment is that people lie about their
> requirements -  as they do, regularly.

Which is why I say to drop it completely. Don't ask for anything of the
sort.

> The bottom line here is that unlike ipv4, there is no practical scarcity
> of ipv6 address space.

The problem is that there is an expected shortage in 'routing table
size'. For the rest there is no other problem.


> The statement that you refer to means in
> practice: "as a LIR, you are entitled to an ipv6 address block, so long
> as you tell us that you have plans to do something, at some stage".

So here you want to infer something from a shorter bit of text again?
While with the above sentence you want to keep it to the letter and
claim that it is everywhere?

Reading texts the way you like it at the moment that you want it is not
what is meant with the text. You are not okay with ignoring the 200
issue, which is only a plan, and who doesn't have a plan, but you do
want other text to be changed because you can't read over it!?

> Is this a bad thing?  Are you really saying that RIPE should create
> barriers to allocating ipv6 space to LIRs?
> If you want barriers, could you please explain why?

I have never mentioned such a thing. Please quote me if and where I did.

Short summary of my vision:
 - Anyone who can pay $fee to RIR should be able to get a /48 or larger.
 - Size depends on what needs they have.

and that is all. The fee is still there mainly there so that not every
Joe Smuck goes around getting address space, a little barrier (monopoly
anyone :) There is no way around that. Clearly there are people who
currently complain already that LIR fees are too high, while their
bandwidth bills are much much much higher and that they can probably
cover those costs with ease with their income. If they can't, they are
most likely in the wrong business.

> Why shouldn't a bona-fide LIR be allocated a first ipv6 /32 as a matter of course?

Because address space should be given out on a 'per-need' basis, not on
a 'you paid 1500 eur a year here have space' kind of deal.

If they only need a single /48 (which currently is the minimum to give
to a site, until the /56 proposal goes through) then they should only ge
t a /48. If they have a huge plan of requiring at one stage way more
address space, then they are planning for at least 200 sites (/32 has
65k of them) and then they should be able to get a /32.

> What are they becoming LIRs for, if not to sub-assign address space?

Most LIR's are LIR so that they have address space for their own
business, not for other businesses. That is, they delegate chunks of
their PA space down to their customers. Which is NOT *PI*, which is what
most people who are complaining at the moment want. PI Is also what ARIN
has given to their members, no change was made in the PA policy.

RIPE folks can probably give the stats on how many LIR's actually give
PI out and how many PA's have been 'sub-assigned'.

> As an interim solution, I'm in favour of this proposal as it stands and
> believe that it should go though.

As I asked Jordi: which type of organizations does this policy help to
get address space and which ones are still left in the cold ?

> We've wrangled about it long enough,
> without making a large amount of headway.  Meanwhile, LIRs are being
> forced to lie to get ipv6 space, which IMHO is plain stupid.

Good that you did not PGP sign that eh ;) What is plain stupid is that
you lied for it. If you are not going to use a /32, or whatever amount
of address space that you received, then why did you lie for it?

What you should have done is made a proposal in which you would not have
to lie. Submit it and then not revoke it.

> However, I would like to see a proposal in the future to remove the
> requirement prohibiting de-aggregation.

Which is what I mentioned what has to be changed, or more specifically,
simply be dropped from the revised text.

> Firstly, routing table
> announcement policies are not RIPE's business.  Secondly, de-aggregation
> and ipv6 table size will simply not be as much of a problem in ipv6 as
> ipv4, because a) there is no supply constraint and therefore no reason
> to split larger blocks

Ever heard of the magical words "Traffic Engineering"? As that is the
cause of lots of /24s being announced, not because so many people have
their small multihomed blocks.

> b) we've come a long way since ipv4 assignment
> started and now have good policies in place which will strongly
> encourage aggregation

Which policies are those?

> and c) most larger operators are - by policy -
> going to filter on RIR allocation boundaries for PA ipv6 space anyway,

This is clearly not the case, look at the current BGP tables.
Both the IPv4 and IPv6 ones btw...

Greets,
 Jeroen


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 311 bytes
Desc: OpenPGP digital signature
URL: </ripe/mail/archives/address-policy-wg/attachments/20070318/b2f342a7/attachment.sig>