Subject: Re: [vserver] Re: Linux source address selection vs. EUI-64
From: Eugen Leitl <eugen@leitl.org>
Date: Tue, 16 Nov 2010 18:56:18 +0100

On Tue, Nov 16, 2010 at 06:13:56PM +0100, Johann Borck wrote:

>> I beg to differ. Dust mote sensor networks can easily be in
>> excess of billions or trillions devices.
>>
>>    
> Hi,
> I followed your discussion and thought above numbers justify a small  
> thought-experiment. I'm not a physician, but you're a chemist, and will  
> therefore be able to correct the mistakes I'll make in the following:  
> According to Wikipedia, world energy consumption in 2008 was 474  

Total solar output flux is ~4 MT/s, Earth intercepts 2 kg/s of that.
We use about 10^-4 of those 2 kg/s, or about 0.1 g/s. 

> exajoules. Now take 2^56 nodes, and you have  0.001827 kWh or 1.827 Wh  

I presume we've dropped the /64 node limitation. Notice that at the
*current* allocation practices we're only a factor of 10^6 to 10^9 bigger
than the current IPv4 space. With your 10^16 you're already by a
factor of 10^10 to 10^7 off.

> or 6578 J energy per node per *year*, ~ 200 uJ/s , which is unlikely to  

Erasing one bit takes KTlog 2 of energy, unless the computation is 
reversible -- so there's no limit, though it can get a bit slow.
Communication of course takes also at least a few photons, so it
will be slightly above that.

> be enough even for future dust mote sensor networks. In the real world,  
> with IPv6, the limiting factor isn't address space anymore, it's energy.  
> Or put in more drastic terms, long before we've run out of address  
> space, mankind will be starved or frozen to death, since we'd allocate  
> each and every Joule of "available" energy to some kind of computer 
> system.

Of course, but that limit is 4 MT/s. And yes, it will be all
used eventually.

>> Moreover, the solar system is pretty big. Add self-replicating
>> hardware, and suddenly 128 bits and 35 years don't appear
>> that much.
>>    
> Wow, assuming self replicating hardware, why not also assume self  

Such are not nearly as far-fetched as people seem to think:
http://www.molecularassembler.com/KSRM.htm

You might have heard of current rapid protoping systems a
la RepRap.

> replicating address space, just to put your mind at ease?  ;)

That's just the point, once you've locked in a standard you cannot
change it. Consider the amount of pain the current IPv4 to IPv6 will
entail, including uneasy co-existance of dual-stacked network, tunneled
protocols, autoconfig breakage and other funnies.

Admittedly, I think IPv6 will by that time will become obviously legacy.

> regards, Johann
-- 
Eugen* Leitl <a href="leitlhttp://leitl.org">leitl</a> 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