idea for ipv6 allocation scheme

[John O Comeau]

Just to summarize the responses:

> As if the Privacy people weren't paranoid enough already :)  The MAN
will
> surely know what they are up to then...

The MAN already knows. And I'm as paranoid as any, but the whole purpose
of this is to wire up the 90+% of the world's population who don't care
what the MAN does. I think the MAN might even like the idea enough to foot
the bill.

> Interesting idea but 2 immediate issues come to mind:
> 1. How do I address the 764 devices and processes in my home or office
> or the 140 unique devices in my motor car?
> 2. What if I don't want to give out my address?

Guys like you and me will always be able to get a few gazillion ipv6
addresses using 6to4 or other means. The idea here is to get the rest of
the world online and sharing information and ideas.

> I wouldn't use this scheme for allocating IPv6 addresses, it is the job
of
> DNS to make it easy to address a computer containing your webpage
right? You
> also face the problem of privacy (as someone mentioned), ISPs sharing
> addresses (very large routing tables?), and management costs.

> I don't think it is necessary to start out with the planets name, first
of
> all we are not there yet, secondly I can imagine the lag involved in
> greate distances and a threeway handshake. If we just could send
> information faster than light..

Good points. But I kind of like the idea of eliminating DNS in at least
one addressing scheme. The ISP sharing thing would be a big problem, I was
thinking however of a separate infrastructure, either government-funded or
grassroots-initiative.

I'd still give a byte up for the planet code, for the 'cool' factor if
nothing else.

> Not to mention that I have an IPv6 NETWORK in my house.  Do I have to
> have a IPv6-NAT for that?

Another inveterate hacker. You _could_, but you wouldn't need to, because
you can also get addresses by other methods.

> You forgot to think about the possibility of having more than 1 computer
> per street address.
> What about those 1/4th mile tall skyscrapers, without counting all the
> wall switches aso...

> IPv6 has chosen to keep the last 64bits as a private address part...
> Can you shrink all the above in only 8 bytes ?

Didn't forget that at all. This is not about J. Random Hacker, it's about
normal people (the kind that don't subscribe to lists like this).

And who says we have to play by the rules, anyway? Even if I agree on the
last 64 bits being private, who says I can't subnet those out even to
/128s? And yes, in many cases, with a good compression scheme, I could get
the above into 8 bytes. 2 for the TLA, 1 planet, 1 country, 1
state/province/prefecture etc., 1 city; makes 6. For a small town with
less than 65,000-some-odd addresses that would work. 

> Heh, Really, the idea I think would have promise, would be a
conspirist's
> nightmare ;)

I'm something of a conspiricist myself, but I think speading connectivity
would have more advantages than disadvantages.

> You might want to check out Tony Hain's (additional) Provider
Independent
> assignment idea:

> http://www.ietf.org/internet-drafts/draft-hain-ipv6-pi-addr-00.txt
> http://www.ietf.org/internet-drafts/draft-hain-ipv6-pi-addr-use-00.txt

> This creates a more unique 48-bit prefix and leaves the rest to the
users.

I shall do that. I'm sure there are many ways of doing this.

> I personally think the above is worse than John's idea. Not only does
this
> require that these /48's be globally routed making it near impossible to
> aggregate this address space, but it also gives an attacker the 10m
location of
> a persons or organisations internet access point! A bombers paradise.

> The other problem with John's idea is that the address of a machine is
tied to
> its address. Not only would routers have to contend with new streets and
houses
> being built, but what if you move a lot?

> These problems can be oversome I'm sure, but perhaps it's best discussed
in
> context of the next version of IP (IPv8 anyone?). ;)

You move, you change addresses. That's the point. And if the
infrastructure is laid out well, the impact on routers will be minimal.
 
IPv8 would be great, but I was hoping to do something before I get much
older; I'll bet we'll be facing the time_t crisis before ipv8 gets off the
drawing boards.

> Zip codes are probably a way to go.

> I once thought about network adressing based upon lat and long; the idea
> was that routers could, knowing their own location, use this information
> as a hint to route packets a bit closer (physically) to their
destination
> when network route information was lacking. If you make certain
> constraints on the physical architecture, you can even ditch BGP and do
it
> all with coordinates. Just need to make sure that, if you drew the
network
> as a big diagram, no area bounded by network connections (with no
> connections across it) is convex, or else you have a "peninsula" that
can
> fill up with traffic meant for points beyond it.

> The cost of administering that is probably less than the cost of RIPE et
> al and all those complex BGP implementations and routing tables :-)

Zipcodes would work great in the US, since the USPS already does all the
work of mapping out new addresses to delivery-walk 11-digit codes. An
11-digit decimal number will fit easily into 5 bytes, leaving room for the
TLD and country code. I may have to drop the planet code after all...

Thanks everyone for the comments. Even if the IANA can't be persuaded to
grant a TLA for this crude idea, it could be used by neighborhood LANs
which don't connect to the internet... I'll post a webpage on this if I
make any headway.

jcomeau@world.std.com aka John Otis Lene Comeau
Home page: http://world.std.com/~jcomeau/
Disclaimer: Don't risk anything of value based on free advice.
"Anybody can do the difficult stuff. Call me when it's impossible."