[ppml] Policy without consensus?

David Conrad drc at virtualized.org
Tue Jan 24 02:12:05 EST 2006


Hi,

On Jan 23, 2006, at 6:49 PM, Daniel Golding wrote:
> We may have to change routing paradigms at some point.

Or we could stop pretending the existing routing paradigm can be  
ignored.

> When we approach that
> scaling limit, it can be considered and examined.

In such a scenario, I imagine one concern would be how to deal with  
all the swamp of legacy IPv6 allocations.

> We are busy worrying about
> being able to route the entire IPv6 space.

I really don't think anyone is worried about that.

> The funny thing is, unless there
> is a reasonable allocation policy, IPv6 will end up on the dust  
> heap of
> history.

Not that it matters, but I haven't been convinced that allocation  
policy is what has been holding back IPv6 deployment.

> Some folks assume that enterprises are willing to swallow a lack of  
> PI space
> and multihoming. Wrong - they are not and will not now or in the  
> future.

Well, they have been in the IPv4 world.

> They buy carrier services - and they will not buy IPv6 services  
> without PI
> space and true multihoming.

I might put it differently.  If IPv6 service provided PI/true  
multihoming, something they have difficulty getting in IPv4, they  
might buy IPv6 services.  SHIM6 could, in theory, provide PI/true  
multihoming, although one might argue it is approximately equivalent  
to starting the IPv6 code deployment from scratch.

> If we can't allocate IPv6 space to enterprises, then its time to  
> scrap IPv6
> and start again.

It isn't a question of IPv6 per se, after all, IPv6 is just how (a  
large number of) bits are organized.  The real question is that of  
routing technology and routing policy.  IPv6 uses CIDR and provider- 
based addressing, just like IPv4.  Enterprises are (presumably) leaf  
nodes in the provider-based routing graph.  Allocating PI space to  
enterprises is promoting leaf nodes to the highest level of routing  
graph.  If you want to do flat routing, that is fine, but people will  
complain that it won't scale (and it won't), although where  
scalability bites you is, as always, open to discussion.

> There is really no middle case in the real world.

Sure there is.  In my experience, most businesses generally don't  
have either the time nor the interest for architectural purity,  
religion, or the perfect solution.  They just want stuff to work  
"good enough".

-drc
--------
My opinions are my own and do not necessarily represent the
opinions of any organization I may be a part of.  So there.





More information about the ARIN-PPML mailing list