Hi Katsuyasu,
My response to Yi was also intended to cover your same concerns about
routing table explosion, but I'd like to add some further thoughts:
I suspect that there is no hard data available and too many variables
involved to produce any real modelling on this, so I think that it is
ultimately down to individual opinion as to whether one believes a
routing explosion would be likely or not. However, some of my reasons
for believing that any resultant growth in routing would be manageable include:
- IPv4 routes have only reached 400,000 to date, and the IPv4 system
started with much more fragmentation than we have with IPv6, and
there are many end sites advertising multiple IPv4 address ranges,
whereas I think the average route advertisements per IPv6 end site
will be much closer to 1.0. So my personal guess is that the worst
case - and over a very long period - would be for the IPv6 table to
have a similar scale to the IPv4 table, and that actually it will
probably be better.
- Core routers now claim they can handle 2 million routes or more,
and I suspect that there will be continued increases in scale; so
while I wouldn't want to encourage more routes than absolutely
needed, it would appear we could as an industry handle triple the
number of routes (IPv4 + IPv6) we are handling now, or even more, if
it were necessary.
- A provider independent (PI) range would only be useful to a
customer subscribed to an ISP service which supported the use of PI space;
- And again the price of subscribing to an RIR to obtain a provider
independent range would probably be higher than most would pay unless
they really needed to.
I realise that you were actually seeking other opinions from the
community besides just my own; I'm hoping that this response may
assist in prompting some further comment from others on the list as
to perceptions on the risk of routing explosion. But I would value
any thoughts you have, whether that be commenting on some of the
aspects I've mentioned above, or raising other reasons you feel might
increase or reduce that risk if we removed the multihoming requirement.
With best regards,
David Woodgate
At 02:51 PM 19/01/2012, Katsuyasu Toyama wrote:
Hi David,
I proposed IPv6 portable assignment, got consensus here, and it was
implemented a few years ago. At that time there were no IPv6 portable
assignment for enterprises or small companies who needed multihoming
just like in IPv4.
The biggest concern of this community was explosion of routing table.
But people agreed the needs for multihoming, and allowed portable
assignment only for it. Renumbering issue and independence from ISPs
were also discussed, but they were not justified for portable assignment.
If routing table explosion is now solved technically or becomes
negligible, the multihoming requirement may be removed.
I also would like to hear voices of this community here, as I think the
concern still exists.
Thanks,
Katsuyasu
Katsuyasu Toyama
JPNAP by INTERNET MULTIFEED CO.