Re: *LA questions [was Re: initial issues]

[Brian E Carpenter <brian@hursley.ibm.com>]

The thing that may be dead is not provider-based allocation, but
the idea that providers will be interconnected in a hierarchy
rather than just a flat space.

To be clear, the original idea was that there would be flat routing
among the (max.) 2**13 providers with TLA prefixes, and all other 
providers would hang off one of those, and therefore their prefixes 
would never be seen in the flat part (the DFZ). Obviously we never 
expected that to be 100% true, but what Randy seems to say is that 
it won't be true to any significant extent, and anything calling 
itself a provider will insert at least one prefix in the DFZ.
That takes away the 2**13 limit.  

  Brian

> Nigel Bragg wrote:
> 
> Please indulge a newcomer to this debate, but this issue seems to be,
> to put it mildly, fundamental.  You can only achieve address aggregation,
> thus control of the number of routes advertised in the DFZ, by alignment
> of address allocation with network topology;  hence provider allocation.
> 
> Achieving this aggregation with multi-homed sites is not entirely
> straightforward, hence the "IPv6 multihoming" issue in the first place.
> 
> If however provider allocation is dead,  the route load imposed by
> multi-homing seems likely to be merely a modest incremental on the
> overall DFZ route load imposed by an ever increasing number of FCFS
> allocations.
> 
> Or have I missed something  ...
> 
> Nigel
> 
>      -----Original Message-----
>      From:   Brian E Carpenter [SMTP:brian@hursley.ibm.com]
>      Sent:   Friday, February 16, 2001 4:36 PM
>      To:     Randy Bush
>      Cc:     itojun@iijlab.net; Sean Doran; multi6@ops.ietf.org
>      Subject:        *LA questions [was Re: initial issues]
> 
>      Randy Bush wrote:
>      >
>      > > - a SLA, or a leaf site (/48)
>      > > - an NLA, or small ISP (/n, where n < 48)
>      > > - a TLA, or big ISP (/16, /29-35 sTLA, or /24-28 pTLA)
>      >
>      > please remove *LA terms.  they are dead.
>      >
>      > the operational community learned long ago that, for many reasons, we can
>      > not say small/medium/large site/isp/registry.  we start allocations with
>      > smaller windows and increase the window size based on actual utilization.
> 
>      Randy, since I haven't seen any discussion of this on the ipngwg list
>      I will ask my two questions here - it seems to be key to multi6, although
>      the proposal to drop the *LA designations is a mainstream ipngwg issue
>      that I assume we will soon be debating there.
> 
>      1. Are you asserting that concept of a site is dead? If not, then SLA
>      and /48 is definitely not dead. (If yes, then we have a big problem
>      but I won't argue it here and now.)
> 
>      2. Are you asserting that concept of a hierarchy of providers is dead?
>      In other words do you think that the idea that lay behind the TLA/NLA
>      split - that local providers hang off less local providers - is totally
>      broken, or is it merely the definition of fixed boundaries in the
>      TLA..NLA portion of the address that is viewed as a problem? Note, I'm
>      not hung up at all on the terminology of "top level" and "next level".
>      I'm just trying to understand what you think will happen in the
>      topology.
> 
>        Brian