[rrg] RACHH: the host-based solution - prepping the PR team to sell it

[Iljitsch van Beijnum]

On 25 nov 2008, at 5:04, Robin Whittle wrote:

> I don't understand this clearly.  IPv4 isn't "legacy" and won't be
> until

Semantics. IPv4 is nearly 30 years old and severely behind the times.  
Even its replacement, IPv6, isn't very new. The fact that many people  
are running IPv4 doesn't change this.

> Even if IPv4 applications are in the minority, and
> still need to run, I don't understand your reference to 14 million
> /24 prefixes.

Current practice is that you need a /24 to gain entry into the world  
wide routing system. We have 221 usable /8s * 65536 = a maximum IPv4  
routing table of 14483456 entries. That is of course 50 times as big  
as what we have today, but at least there is some kind of reasonable  
limitation here.

>> Hm, I'd rather redownload 5 out of 10 files when my ISP craps out  
>> than
>> 10 out of 10. Partial deployment for multihoming is still useful.

> Yes, but I would say that for any substantial organisation, running
> servers or having a bunch of desktop users or both, that they are not
> going to invest in some new technology, more complexity, new
> addressing arrangements perhaps and a second ISP if the new
> multihoming scheme is only going to work for half their traffic.

We have less than 25000 multihomers world wide today, so apparently  
most people are happy with even a single ISP, and would presumably be  
even happier with some form of multihoming even if it doesn't provide  
the full benefits of the "real" thing with PI addresing, especially if  
there are no additional costs beyond a second connection.