Re: [vserver] Re: Linux source address selection vs. EUI-64

From: Eugen Leitl <eugen_at_leitl.org>
Date: Mon 15 Nov 2010 - 20:05:37 GMT
Message-ID: <20101115200536.GD28998@leitl.org>

On Mon, Nov 15, 2010 at 06:59:59PM +0100, Romain Riviere wrote:
> Good evening,
>
> This is interesting. You have just succeeded in
> turning what (I hope) was a friendly exchange of opinions into a patronizing rebuff.

Very sorry if you perceive it that way. No such intention.
 
> What is most interesting is that I believe we agree
> that IPv6 address space has to be carefully allocated.

Very much so.

> We simply appear to disagree on how careful we judge
> existing allocations so far. And I am beginning to see why.
>
> Le 15 nov. 2010 à 10:45, Eugen Leitl a écrit :
>
> > On Sun, Nov 14, 2010 at 02:22:40PM +0100, Romain Riviere wrote:
> >
> >> All I'm saying is, for the time being, catastrophic
> >> predictions that we're making the same mistakes as
> >> with IPv4 are a little dramatic and possibly premature.
> >
> > Strange, I provided you with evidence that IPv6 space
> > will run out within about the same time as IPv4, using
> > pretty conservative assumptions (which we know won't
> > hold).
>
> Evidence, no, unless you know the future.

I don't, of course. It's just from the current
state of allocations and certain built-in assumptions
the effective space is only a factor of millions to
billions bigger of IPv4. Since we're at the beginning
of IPv6 allocation, it could well be that the space
runs out in roughly the same time as IPv4 did, it
offering only 4 gigadresses total and being empty
initially.

> Constructive arguments, yes, I'll give you that.
> But please do not mistake opinions for facts.

The problem is that we have two outcomes, one where
the cassandras are wrong, and one where they are right.

If the cassandras are wrong, and all is dandy, we
don't have to do anything, which is great.

However, if the cassandras are right, we have to deal with a
yet another messy transition, which is avoidable since
could be postponed for a much longer time.
 
> You are therefore suggesting that public IPv6 address
> space be allocated to motes and other nano-machines.

Yes, that is correct. I also think the cut-through purely
photonic routers will pretty much force us to encoding
geographic coordinates into packet headers, with
that scheme being utilized by routers everywhere.
No more global routing tables, just a list of local
deviation of the local hyperlattice from idealicity,
with reliable proximity metric to packet target
available everywhere.

Potentially we're looking at at least ~um-scale
resolution on Earth surface,
with bigger voxels as you go up. You'll probably have
to define a hierarchical addressing space for other
gravity wells, however minor or major, as long as they are big
enough to reliably bind nodes in orbits.

Prior address assignment domains throw a major monkeywrench
into that scheme. Very possibly that IPv6 (assuming we
will really see major adoption, the current rate of progress is
quite glacial, despite the symbolic "almost running
on empty" signs) will have to be completely scrapped
not very many decades from now, and will co-exist as
a tunnelled protocol, presumably how IPv4 will suffer
that fate not so very many years from now.

> Is this a fact or an assumption ? Why not link-local
> for example ? And please accept this as an honest question, it really is.

You can probably also address individual swarms or clouds
hierarchically, but they're going to need a way of routing
within themselves, including overlap with other systems.
For efficiency sake you want your entire swarm volume to
route packets, and not just use traffic peering at a few points.
I don't think you could use NAT here, if you have to encode
geographic position, which can be fundamentally open-ended.
 
> In short, this is the heart of our disagreement, or
> so it seems. If tomorrow's nano-swarms *have* to c
> ommunicate using *public* space, then yes, we're
> already doomed and all hail IPv8/16/42.

> I had honestly never considered this before, point taken.

With that I've pretty much succeeded with what I wanted to say.
 
> > in the solar system.
>
> Then we have to bear in mind that IPv6 simply might not cope with that.

Point taken.
 
> The mass idea was just a (poor?) joke: if we have
> enough address space for every molecule in the Earth

As a chemist, I like to point out to people that we
might see single individual systems containing roughly
Avogardo numbers of switches (10^26 bits) within
our lifetimes. Even assuming some 10^4
atoms/switch that'd be only 120 kg of carbon.
A cubic micron of such switches really goes a long
way.

> crust for example, we should be safe, until we start
> creating stuff out of matter from the mantle or
> elsewhere. But I have to admit that putting public
> address space into nano-machines make it a completely different business.

Of course talking like that makes one sound like
a lunatic, and hence easily dismissible.
 
> >> Until we create matter out of nothing, I still
> >> think we're on the safe side. I don't doubt
> >
> > I'm pretty sure that's exactly what early
> > ARPAnet people said when they said 32 bit is going
> > to be enough for everybody. I mean, how can
> > computers be owned by individual people, never
> > mind *multiple* computers?
>
> 640k should be enough, etc. I hear ya :-)
>
> >> you disagree, and I'm ok with it :-)
> >
> > It's too bad I can't see you eating your hat 35 years
> > hence :)
>
> Have we entered a contest that I am not aware of ?

No, just a jokular term for eating one's hat, or
more like eating crow.

> I am really OK with differences in opinions.
> Is it really that hard to believe ?
> Again, let's not make this personal, shall we ?
> Because I can't see a reason why it should be and why I should participate in that :-)

Again, this is meant to be completely amicable, and as you
know it is very easy to interpret emails as something
they're not intended to be. Apologies for anything
resembling lecturing.
 
> Cheers,
>
> Romain--
Eugen* Leitl leitl http://leitl.org
______________________________________________________________
ICBM: 48.07100, 11.36820 http://www.ativel.com http://postbiota.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A 7779 75B0 2443 8B29 F6BE
Received on Mon Nov 15 20:06:24 2010

[Next/Previous Months] [Main vserver Project Homepage] [Howto Subscribe/Unsubscribe] [Paul Sladen's vserver stuff]
Generated on Mon 15 Nov 2010 - 20:06:25 GMT by hypermail 2.1.8