[ppml] Suggestion for ARIN to deligate smaller IP blocks

[Paul_Vixie at isc.org]

> > > The magic words, "private network", appeared in the next paragraph
> > > you snipped, don't remember if I made it clear I was talking about
> > > private internets, not intranets.
> > 
> > what's a "private" internet?
> 
> I mean an internet (a network of networks) between different
> organizations, but not (directly) accessible to the Internet.

that's far from definitive.  first off, by "the Internet" you seem to mean
what some people call the default free zone (DFZ) but you might be using
breidbart's definition, which i've quoted a time or two here recently but
here it is again:

	>> But what *IS* the internet?
	> It's the largest equivalence class in the reflexive transitive
	> symmetric closure of the relationship "can be reached by an IP
	> packet from".		--Seth Breidbart

second and more importantly, if i'm a member of more than one of these
"private internets," say one between "different organzations" A,B,C,D and
one between "different organizations" D,E,F,G where i am "D", what am i?

third and finally, if the rest of the world nukes itself in spam and ddos
and "different organizations" A,B,C,D,E,F,G decide to form their own
"the internet", or are the last networks standing after the spamocaust
such that we qualify under breidbart's definition, then what am i?

> Much has been made of the definition of "the Internet" as the largest
> set of host defined by the relation "reachable by each other" (as I
> understand  it) in some recent posts.  There are many other internets,
> networks encompassing cooperating entities, which contain hosts that
> are not reachable from the Internet at large but are reachable by each
> other, and which have some hosts that are part of "the Internet."

i agree that any network of networks is "an internet".  it remains to be
seen whether your use of the term "the internet" reaches the same concept
in my head as it does in yours.  word definitions may not matter if we
simply avoid disputes and ambiguity.

a network is "private" if it's not connected to most other networks, but
clearly it could still be connected to some other networks and still be
"private".  while the U in ULA ("unique") is clearly beneficial since a
network might be less private at some parts of its life cycle than at
others, and collisions are painful, the L in ULA ("local") is unclearly
unbeneficial since during the times in its life cycle when a network is
less private, it might be painful to have someone refuse a third party
route on the basis of its policywise L-ness.

> They need non-colliding addresses.

all RIR allocations (PI or PA) are non-colliding ("unique"), so this part
is already handled.

> RFC1918 might work fine, if a single person or group were managing address
> assignments (and the private internet were small enough), but I'm talking
> about intERnets, not intRAnets, that is, these networks belong to and are
> managed by different people, groups, companies, or organizations, so RFC1918
> addresses are very likely to collide.

i'm not proposing that you use RFC1918 or IPV6 "site local" for this kind
of network.

> Assigning unique addresses to all the involved organizations would prevent
> the collisions.

agreed.  for example, PI or PA.

> I realize that given the pending crunch in ipv4 addresses, it may not be
> possible/practical to assign even /24's to all organizations who might need
> them, but there is *no excuse* for not allowing for this in ipv6.

also agreed.

> Whether it is ULA-C or PI is immaterial.

here, i disagree.  ULA would be a waste of address space, and administrative
effort.  there's more space available for PA and PI than will ever fit in the
DFZ.  your argument supports a policy for smaller-size PI allocations which
may be cheaper and may be easier to qualify for and may be allocated out of
a well known /10 to make them safe from static TE-resistant route filters,
but no argument i've seen here supports a policy for "unique local addresses".