[ppml] Draft ARIN Recomendation

[bmanning at vacation.karoshi.com]

On Thu, Oct 21, 2004 at 10:38:42PM -0400, Thomas Narten wrote:
> Hi Geoff.
> 
> Some good points to think about.
> 
> One regret I have about the meeting is not asking for a "sense of the
> room" when this topic came up. It would have been really useful to get
> a better sense of how widely shared the view is.
> 
> Also, since no one has posted the 'problem' yet, let me give a summary
> of what I thought I heard during the meeting:
> 
> The jist of the issue as I understand it is that folk are worried that
> there will be tremendous pressure on ISPs to route ULAs and that they
> will in fact become the PI addresses in practice that we don't know
> how to route in a scalable fashion. And at some point in the future
> we'll have a huge mess with only ULA addresses being used. I.e., folk
> (end users) will see these as PI addresses, won't notice the fine
> print about "not globally routable" and will force ISPs to route
> them. And since this works in the short term, at the point where it
> becomes a real problem expectations will have been sent and we can't
> roll things back.
> 
> Is this even close to what people are thinking?
> 
> Thomas

	the jist of this debate lies, for me, in the presumptions
	about the term "globally" - esp.when it appears next to 
	routable.  in the exiting miasma we call "Internet" there 
	are -very- few prefixes that are constrained. I made a short
	list at the mic and will reiterate here:

		0.0.0.0/32
		255.255.255.255/32

		and the vendor constrained 

		127.0.0.0/8

	-EVERYTHING- else is routeable.  We had to fix that when
	CIDR was deployed by getting vendors to remove the "martian"
	and "guardband" prefixes so that cidr blocks could be used.

	If a prefix is routable ... its routable (duh).  Global
	is a novel idea... since the days just after the MILnet split,
	and the emergence of EGP protocols, policy has been an implicit
	component of routing.  A network is allowed and implicitly encouraged
	to establish a policy on what prefixes it will accept. e.g. 
	some prefixes are or will be unacceptable and packets destined
	for those prefixes will be dropped at the boarder.  Hence
	the illusion of "global" routing scope is a myth.

	The promise of ULAs - invites chaos and anarchy.  The allure 
	is near zero aquisition cost, near zero accountability, and for
	early movers, near zero insertion cost.  These will be treated
	as PI space (and I hate that term) and can -NEVER- be eradicated
	once released.   (sez bill, donning his nostradomus robes and hat)
	Perhaps one might bounce the ULA idea off our local (US) LEA folks...
	I'm sure they would -love- the idea of anonymous delegations.. :)

	so yes, you pretty much nailed the point.  There are a host of others
	but lets work on this one for now.

--bill