ietf
[Top] [All Lists]

Re: IPv6 will never fly: ARIN continues to kill it

2007-09-13 05:57:47
Dear Jari;

On Sep 13, 2007, at 4:24 AM, Jari Arkko wrote:

Michael,

Here's a decision table for you:

1. Do you need addresses that are routable from the global
    Internet, from anywhere?

    (Its not clear to me that you do, because you only need to
    do that within your own network and a couple of well
    known external sites perhaps.)

    a. If not, maybe you should look at ULAs. RFC 4193 allows
        you to get these addresses randomly, and you do not
        need to ask permission from anyone to do it. You could
        have your addresses today if you wanted to.

    b. Proposals have been floated about non-random ULAs
        as well. Right now we do not have one, but I'm not
        sure you need this for your particular case.

2. If you do need addresses that are routable, is it
    sufficient for you to work with provider-aggregated
    addresses that you get from your ISP (not from ARIN)?

    If yes, get the addresses and use them!

3. If you do need addresses that are routable AND
    you have multiple ISP connections and want to stay
    away from an address renumbering if you need
    to change ISPs, then you need PI. You are starting
    to get PI space, but as numerous PI items in the
    global routing table cause pain for routers, this
    will likely be available only for larger enterprises.


I do not really agree with this. First, the routing tables do not care if you have PI or PA space, just whether it is announced or not. If you are already announcing PA space, and getting into the DFZ, it does not harm the tables if you change to PI space. Second, one of reasons I helped to write and push through ARIN 2002-3 (micro assignments) was that I felt that small multi-homers (i.e., enterprises that were multi-homed but did not need large address allocations) did not constitute a threat to the routing tables, and that has been borne out by experience. Neither the growth nor the bloat in the routing tables is being driven by small multi- homers. This has been discussed at great length on ARIN PPML and other lists.

Yes, I gave numbers to Vince Fuller about millions of multi-homers, but that was to set an upper bound on the process. I do no believe that every small business will rush out and multi-home, no matter how automated BGP becomes. The small businesses that I know that multi- home (mostly high traffic volume companies providing network services, such as video streaming) have a business need to do so, and it is not realistic nor in my opinion proper to assume that they will not be able to do so, one way or the other.


    There is ongoing work to try to design a better
    routing system that would be capable of keeping
    tens of millions of prefixes or more, in the IRTF.
    If and when that work succeeds, it would be possible
    to allocate everyone their own PI prefix. We are
    not there yet.

In any case, FWIW, I think it would make sense for RIR
address allocation rules to allow IPv6-only operations
and not just those that need both IPv4 and IPv6 address
space.

I fully agree here. In fact, I would say that IPv6 will have truly succeeded when business requests start coming in that do _not_ want IPv4 space. This should be encouraged, not discouraged.


Jari

Regards
Marshall



_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf


_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf