[GLOBAL-V6] Re: [GLOBAL-V6] Re: [GLOBAL-V6] Re: [GLOBAL-V6] Re: Válasz: Re: V álasz : [ GLOBAL-V6 ] RE: How to reduce the junk ap plications?

[Gert Doering <gert@Space.Net>]

Hi Steve,

On Tue, Feb 19, 2002 at 03:58:22PM -0800, Steve Deering wrote:
> At 12:28 AM +0100 2/15/02, Gert Doering wrote:
> >I know what the draft says here, but I still claim that "number of used 
> >/48s is not a useful metric to judge a network's size (or worthyness of 
> >anything)".  Neither is "number of used IPv6 addresses".
> 
> Assuming the policy of one /48 per end-subscriber site is adhered to,
> the meaningful metric -- *for the purpose of addressing allocation* --
> is the number of end-subscriber sites (regardless of size or "worthiness")
> served by an ISP.  

Ummm, I was unclear.  What I want to say is "the number of /48s is a 
meaningless metric to decide whether someone should get his own 
'independent' IPv6 address block or not".

Of course it is meaningful to calculate the *size* of the block, if the
decision "yes, he will get one" has been made.  This part was so obvious
to me that I didn't mention it.

All my arguments in this discussion always focus on the "you have to
document <n> customers to be worthy of getting your own IPv6 /32" issue. 
The remainder of the draft proposal is fine with me.

> The more subscribers, the more /48s the ISP needs,
> in order to provide distinct routing to each of them.  It doesn't cost
> the ISP any more routing table space or routing protocol overhead to
> route to a big/important site than to a small/unimportant site, so why
> do you think a network's size or importance is relevant in this context?

This comes from "will this organization get their own network block?".

As there are voices that argue that "not ever network (that asks for it
and provides /48s to third parties) should get their own network block",
I tried to point out that *if* we want to impose criteria to judge
whether someone is "big" or "important" or "whatever" enough to warrant
their own network block, the sheer number of /48s assigned is NOT a
useful metric.

I'm not the one argueing for "some should get their block, others should
not" - I'm the one argueing "if you want this, please define *who* should 
get one, and provide criteria that actually work".

[..]
> If you need an "importance" metric (e.g., to filter a routing table that
> has grown beyond supportable limits), go ahead and invent one, but please
> don't try to overload it on the allocation size / prefix length -- we have
> enough problems with deciding allocation sizes as it is.

Can't agree more.

Gert Doering
        -- NetMaster
-- 
Total number of prefixes smaller than registry allocations:   71809 (71887)

SpaceNet AG                 Mail: netmaster@Space.Net
Joseph-Dollinger-Bogen 14   Tel : +49-89-32356-0
80807 Muenchen              Fax : +49-89-32356-299

-
- This list (global-v6) is handled by majordomo@lists.apnic.net