On Mon, Sep 04, 2000 at 06:19:11PM +0000, Alex Barclay wrote:
> On Sun, 3 Sep 2000, Ken Hornstein wrote:
> I came to the original conculsion that renumbering would be easier. In the
> last couple of years this seems to have been downplayed significantly.
> 
> Thinking about it I feel there are two reasons. Firstly it's a damn hard
> problem. Secondly the ipng folks tend to be equipment vendors or network
> operators; equipment vendors don't care directly about renumbering so long
> as they shift boxes; network operators don't need renumbering as it would
> allow end users to change isp's more easily, I could even think that there
> was a conspiracy here :-) 
>
Oh Oh! Do we need to fear for  another hollywood conspiracy movie 
looming on the horizon? I agree with your first reason but the second is 
a bit farfetched I think. I think the ipng folks have done a great job in
supporting renumbering in IPv6, although some things are still under
development (like site-local addressing architecture). The reason it
is a hard problem is that it just cannot be solved by the IPv6
protocol alone. It is to a great extent a network management problem
which in addition to network-layer facilities requires upper-layer
support, like DNS support, network management protocols and tools,
better APIs, even better network security mechanisms and whatnot.

These parts required to make renumbering relatively easy just fall
mostly outside the scope of the ipng working group, and need to be
dealt with in other existing groups or in new working groups that will
spin-off off the ipng wg. Renumbering *needs* be easier in IPv6. It
would be nice for the current Internet, although many big players in
the current Internet are not to concerned as they are quite comfortable
with their own address (provider independent) address space, and
many others have no serious problem with NAT technology which they
also use for other purposes (like hiding their internal network 
structure). 

Easier renumbering is almost a requirement for new big players and
applications on the horizon. The alternative would create just to much
mind-boggling (management) complexity. The Best Practice (or most
widespread) in the current Internet is just not good enough. 

I am currently assisting as a consultant setting up a broadband ISP
(24 hour bi-directional 10Mbs and maybe 100Mbs or more in the far future, 
i.e.  a few years from now, so no pathetic ADSL or shared-bandwidth
Cable ;), requiring basically all the bandwidth current and future
technology can provide in its core network. NAT a feasible option in
such a network? NATPT? I don't think so. Why? Simply, because Cisco
(or any for that matter) doesn't sell equipment that can do such a
thing at multi-Gbs speed. And even if they did it would be
outragiously expensive. And this type of network is growing at an
incredible rate in Europe, requiring massive amount of static or
semi-static IP-addresses from RIPE. Enough to make RIPE quite nervous
and let alarm-bells sound all over. IPv6 deployment 5 years from now?
I believe RIPE believes now  we don't have that much time, i.e. if 
we don't want the growth of the Internet in Europe come to a grinding 
halt. Believe it or not, but the world outside the USA is quite big.

They jumped on the Internet technology bandwagon a bit late, making it
a bit trickier to get that couple of billion IP-addresses they need in
the next few years. We've got about half a billion mobile GSM phones
in Europe and Asia, waiting to get IP-connectivity. That's about twice 
the amount of Internet hosts right now. Why wasn't GSM based on IP
technology? (would be great for data communication, GSM-data is just a 
plain nightmare right now). Many reasons, but reason number one: no
spare address space in IPv4 and IPv6 development had only just begun at 
the time GSM was developed. Now 3rd generation mobile technology is being
developed and will see wide deployment within 2 or 3 years. This will
be exclusivily be based on IPv6 technology. Looking at the success of
GSM over the past 10 years, it is not entirely unreasonable to
estimate half a billion to a billion mobile IPv6 hosts 5 to 10 years
from now (whether this time the USA jumps on this mobile boat too or not).
IPv4 technology with or without NAT just isn't going to do
this. These mobile terminals will need a globally unique identifier
(telephone number equivalent?) and a location identifier, both can
nicely be provided with IPv6 addressing (64 bit, network prefix and 
interface identifier). Renumbering (of the network prefix) these 
machines will be much more common and part of normal operation (i.e. 
not be the  exceptional disastrous event) as these mobile terminals roam
between  different mobile networks (e.g. when you travel outside your
own country or to an area your normal provider has no coverage. Ok,
the latter is pretty rare for most providers here, but still.). Such
roaming just works for GSM (you can drive from the north of Norway to
the south of Italy, talking on your phone without high likeliness of
loosing your connection at any time), and users will expect this to
just work for an IP-based next-generation GSM too.

Just two realistic examples. I'm not even mentioning the  billion or
so currently unconnected Chinese or Africans or whatever other group
might increase their demand on address space. Ok, I just did...

My point? Stating that the grand old big players of the IPv4 Internet
have no great motivation to move to IPv6 technology implies IPv6 is
dead or irrelevant is just plain silly and irrelevent seen from this
side of the ocean. I'm pretty sure that for exactly the same reasons
there are still quite big SNA and DECnet networks around that never
felt the need to move to IP-technology. You don't see the need to pay
the (admittingly high) cost of moving your present networks to IPv6?
Fine, no problem! Just don't get in the way of the rest of the world
while they are building on the required technology for their future,
while you guard your past... Just remember, the future is generally
bigger than the present and the past together ;-).

> hasn't happened yet (especially given my contract for t1 service is for
> multiple years).
I'm sorry to hear that you're stuck with a t1 for the years to come ;-}.

Feico Dillema.