Allocation of IP Addresses

Hi,

Here's an idea. Let new ISP's reserve large blocks (say /16's) in 65/8,
66/8, .... but don't let them actually use these addresses on the global
Internet. Then, the ISP can run a Network Address Translation gateway and

Why not use net 10 and leave the NATs in?

Of course, there is one little problem with this....
bash$ whois 65
Air Force Logistics Command (ASN-LOGNET) LOGNET-AS 65
bash$ whois 96
Army Finance and Accounting Office (ASN-JTELS) JTELS-BEN1-AS 96
How did these guys get such big chunks of address space reserved?

You're probably gonna regret that one: you're looking at AS numbers, not
networks. Now, let see how many people tell you the same thing :-).

I think that the fundamental problem here is that the Internic is
fundamentally clueless about important issues such as global routing

Bullshit. The InterNIC is very much aware of global routing issues.

and *BUSINESS* issues.

What business issues are you talking about?

They are behaving a lot like a government bureaucracy
or a regulatory agency.

The registries are simply following the policy as defined by the
Internet community at large. Read RFC 1466, 1814, and the latest
internet draft on allocation policy. If the Internet community were
to define new policies, the registries would implement those as well.
Currently, there is a small bit of contradiction in what small ISPs
and end users want and what the larger ISPs (and what is necessary to
keep the Internet from partitioning) want. But, both sides are more
than happy to scream and whine at the registries for not doing the
"right" thing.

I don't really see how this can be fixed with the current system of
having a US government agency writing a contract with a private US
company to provide a fundamental international infrastructure service!

The US government agency could get out of the way, but the squeals of
outrage when InterNIC started charging US $50/year for domain name
registrations leads me to believe it will be a while before any sort of
rational allocation policy can be imposed.

Regards,
-drc

>Here's an idea. Let new ISP's reserve large blocks (say /16's) in 65/8,
>66/8, .... but don't let them actually use these addresses on the global
>Internet. Then, the ISP can run a Network Address Translation gateway and

Why not use net 10 and leave the NATs in?

Indeed! RFC1918 addresses work fine for me on my LAN at home and this
week I will be connecting a corporate LAN also using RFC1918 addresses
behind a single static IP address. In both cases I am using a FreeBSD box
with proxies like CERN httpd and TIS Firewall Toolkit to transfer the
traffic rather than a full-blown NAT.

More ISP's should be doing this IMHO.

>I think that the fundamental problem here is that the Internic is
>fundamentally clueless about important issues such as global routing

Bullshit. The InterNIC is very much aware of global routing issues.

Then why have they not yet come up with a workable policy like the one
RIPE uses to release /16 blocks incrementally to new ISP's?

>and *BUSINESS* issues.

What business issues are you talking about?

Basically, the market demand is INCREDIBLY HIGH and businesses want to
build up infrastructure to meet this demand but the Internic IP address
allocation procedures are too confusing and take too long.

keep the Internet from partitioning) want. But, both sides are more
than happy to scream and whine at the registries for not doing the
"right" thing.

I wonder if the main problem isn't simply that not enough people know how
to have an impact on Internet policy as expressed in the RFC's. A heck of
a lot of people starting ISP's have backgrounds in business, LAN's and BBS
operation but they just don't know how the Internet works or where they
can comment on Internet policy. *sigh*

The US government agency could get out of the way, but the squeals of
outrage when InterNIC started charging US $50/year for domain name
registrations leads me to believe it will be a while before any sort of
rational allocation policy can be imposed.

Those squeals disappeared darn fast!

Michael Dillon Voice: +1-604-546-8022
Memra Software Inc. Fax: +1-604-546-3049
http://www.memra.com E-mail: michael@memra.com

Well, currently InterNIC is the registry of last resort, and given RFC
1814 and such, they CANNOT, if you fight hard enough, turn you away.
Given the number of new ISPs that come to them for provider independent
addresses, they isn't enough IPv4 address space to do the above with.

-dorian

Are you sure of this? Even if they start allocating out of the former
Class A space?

After all, getting a reserved /16 out of the former Class A space
wouldn't exactly be free because you would need to buy a NAT in order to
avoid renumbering down the road so not *ALL* ISP's are going to demand
one of these. And it doesn't hurt to publicize the existence of NAT
technology either, because if ISP's know that NAT's exist they are more
likely to deploy them at customer sites along with RFC1918 addresses.

BTW, that puddle on the floor is from the egg dripping off of my face.
I'll be more RIGOUROUS next time at readin the whois output.

Michael Dillon Voice: +1-604-546-8022
Memra Software Inc. Fax: +1-604-546-3049
http://www.memra.com E-mail: michael@memra.com

But that's not what you said. Given that Internic gets about 50-60
address requests a week, if you reserve /16 for each, you can do the
math. I guess this would force the deployment of IPv6 much sooner than
currently projected.

-dorian

> Are you sure of this? Even if they start allocating out of the former
> Class A space?

But that's not what you said. Given that Internic gets about 50-60
address requests a week, if you reserve /16 for each, you can do the
math.

Math? What does math have to do with it? *grin*
Oh well, I guess there's some math there but I find it hard to believe
that out of 50 address requests per week from ISP's there are more than
10 that are truly clueful ISP's that absolutely need a /16 one year down
the road.

There are still a *LOT* of people that think the Internic hands out
portable addresses and startup ISP's are not much different from the
general population in that regard.

Besides, the process of applying for an IP block is a lot like a
negotiation. The newbie ISP presents their case as to why they know they
will be mega-ISP Inc. a year down the road and absolutely must have that
/16 block from the Internic. Then the Internic can counter and ask why
they can't just use RFC 1918 addresses and run a NAT on their gateway.
The ISP can plead poverty that they cannot afford to buy a PIX from Cisco
and then the Internic can point out that a poverty-stricken ISP isn't
likely to need a /16 and counter with a /23. And so on ........

I guess this would force the deployment of IPv6 much sooner than
currently projected.

It is by no means certain that IPv6 will ever be deployed. There are lots
of ways to wring new life out of IPv4.

Michael Dillon Voice: +1-604-546-8022
Memra Software Inc. Fax: +1-604-546-3049
http://www.memra.com E-mail: michael@memra.com

Michael Dillon <michael@memra.com> writes:

  > Then why have they not yet come up with a workable policy like the one
  > RIPE uses to release /16 blocks incrementally to new ISP's?

The RIPE NCC is not doing that.

What we are doing is quite similar to what the InterNIC does with one
very notable exception: We charge ISPs for registration service and we
audit their assignments so they have to have their act together. This
means there are significant resources involved in obtaining address
space from us rather than their transit provider. This causes ISPs to
make much more rational decisions about where to obtain their address
space. It also makes the rate with which new local registries are
established quite predictable which allows for some level of
rationalisation in allocation decisions.

Unfortunately the InterNIC is in no position to put ISPs before that
choice. THIS NEEDS TO BE CHANGED!

Details:

We allocate a fixed size first allocation (currently /19s) to each and
every newly established local registry (ISP) *no matter what their
glorious plans are*. Further allocations are made *exlusivcely* based
on past usage rates which is reasonably rational. Remember that we
audit assignments.

We will do our best to place subsequent allocations such that they can
be aggregated with previous ones. We are reasonably successful at this
and that is probably why the misunderstanding above is quite common.
Formally however we make no guarantee whatsoever about the placement of
allocations. This gives us the possibility to react flexibly and do the
right thing most of the time.

Again, this is essentially the same policy as the InterNIC uses with
a few local variations. The main difference is that ISPs have to expend
resources to get service.

Daniel