[ih] Some Questions over IPv4 Ownership

[John Klensin]

Well, at some level, of course.  The problem here is the hierarchy
which, wrt "ownership" bears more relationship to ownership/possession
delegation in a feudal system than to some notion of absolute
ownership.   Qwest certainly has the "ownership" rights it claims wrt
addresses delegated (or assigned, if one prefers) to its customers.
But its own "ownership" depends on its relationship with ARIN and
ARIN's "ownership" depends on its relationship with IANA.  I'm
disinclined to push back further than that for the reasons given in a
previous note, but others might.

Can Qwest retrieve address space that had been used by a previous
customer? Certainly and it is done all the time.  Can they retrieve or
reassign address space that is being used by a current customer and
force renumbering?   Certainly yes in theory.  In practice, I would
assume that the answer might have more to do with how arbitrary the
decision was and with the pain and costs of renumbering than with
abstract theories of ownership.  And I've seen more than a few ISPs
hesitate to reassign addresses to customers who would find renumbering
sufficiently painful to start hiring lawyers, precisely to avoid
testing the theory of just how far their "ownership" rights extended.

People closer to the current RIR system that I am may want to comment
on this, but I assume the situation is much the same.  I understand
that their current, non-legacy, agreements about assignment/delegation
of address space try to make it clear that such assignments don't
transfer any ownership rights.  In principle, they could take
addresses back from an active user of that address space to whom the
address space had been delegated.  In practice, trying it would
probably be really ill-advised.  And, again, that is less because of
theories of ownership than because renumbering can be expensive and
because the RIRs are consensus organizations and any perception of
significant arbitrary behavior would probably cause a revolution from
their members -- either on the basis of concerns about fairness or out
of concern about who they would come after next.

A different way of looking at all of this is that the entire
addressing allocation/ delegation/ registration model ultimately rests
on a combination of collective consent and an understanding that
having non-unique addresses floating around causes problems in routing
packets.  ISPs have lots of incentives to announce and route only
registered addresses; those who might fail to do so would be
performing a disservice to their customers (unpredictable delivery),
would antagonize their peers and probably be cut off, with the
long-term effect of putting them out of business.   Could someone
appropriate a range of addresses and start using them locally and with
their closest friends?  Sure and before RFC 1918, it used to happen
all the time.  But, as Spencer and others have pointed out, such
networks need to be kept completely isolated from the public network
and its address space; if it is not, we have non-unique addresses and
A Big Mess whether there is a debate about ownership or not.