[ppml] Definition of an (IPv6) End Site

Thus spake "Jason Schiller (schiller at" <jason.schiller at>
> My confusion rests not on who is an end-site, but how many end-sites
> that who is.
> Let me give you an example.  Lets say I am and ISP and I have a customer
> that owns a chain of 100 hardware stores across the US.  Each store is
> independently operated.  Each store has a separate Internet connection
> through me.  Each store has a single LAN that is only connected to the
> Internet.  The stores have no interconnection at all.
> But there is one person at the corporate office that orders all of the
> T1 installs, and all bills are sent to the corporate office under a single
> company name.
> Since there is only one end-user that has a business relationship with me,
> would this only qualify as a single end-site, and thus all 100 locations
> should share a single /48?  Or can I consider each separate network ate
> each separate location an end-site?  In this case I could assign 100 /48s.

Assuming you're talking PA space for this case, there's no harm in giving 
each location a separate /48.  If they planned on interconnecting at a later 
date, I'd suggest they share a single /48.  However, as long as we're 
talking PA space, the definition of "end site" can be up to the LIR.

> What if the sites are interconnected, in that case should it be one
> end-site?

IMHO, yes.

> What if the sites are interconnected, but each site has a different
> contact, but it is all part of the same compnay, in that case should it be
> one end-site?

Yes.  Subsidiaries should only count as "other organizations" if they 
operate an autonomous network with distinct policies and requirements. 
Shell companies, business units, and "internal service providers" do not 
qualify merely because someone draws cute diagrams in Visio.

A wholly- or partially-owned subsidiary is the interesting case.  If someone 
showed me contracts for service, a flow of invoices and payments, etc. I'd 
probably agree they're separate orgs.  If the two had no private 
interconnection (e.g. exchanged email over the Internet), I'd agree.  If 
they shared the same network but had internal "charge-backs", I'd disagree. 
There's a lot of gray in between, however.

> I do understand the other confusion that is going on as well, for example
> how is the IRS an ISP.

IMHO, letting the IRS claim to be an LIR was a clear violation of policy, 
but given that there was no PI policy available, what else was ARIN to do 
when presented with an otherwise reasonable request?  (Not to mention the 
IRS is not a group you want mad at you...)  Ditto for the various other 
state and federal orgs that have managed to get themselves accepted as LIRs. 
However, there's a relatively small number of those, so if it were just them 
it'd not be a dire emergency.

However, that same hole has apparently been abused by enterprises as well, 
and there's a _lot_ of private businesses out there.  Is Cisco really a LIR?
Is IBM?  Only the biggest companies are doing it so far, but those are also 
the folks that will be telling _their customers_ to do the same as a "best 
practice", complete with slideware on how to do it.

> What I want to suggest is if you are trying to tighten the language,
> please also consider the case above .

I believe that's exactly what we're trying to address.


Stephen Sprunk        "Stupid people surround themselves with smart
CCIE #3723           people.  Smart people surround themselves with
K5SSS         smart people who disagree with them."  --Aaron Sorkin