[arin-ppml] Policy Proposal: Open Access To IPv6
gdolley at arpnetworks.com
Sat May 30 21:09:36 EDT 2009
On Sun, May 31, 2009 at 12:45:57AM +0000, bmanning at vacation.karoshi.com wrote:
> On Sat, May 30, 2009 at 05:00:38PM -0700, Garry Dolley wrote:
> > On Sat, May 30, 2009 at 09:28:19PM +0000, bmanning at vacation.karoshi.com wrote:
> > > On Sat, May 30, 2009 at 02:04:22PM -0700, Garry Dolley wrote:
> > > > On Fri, May 29, 2009 at 04:15:02PM -0400, Leo Bicknell wrote:
> > > > > To that end, I can't support the proposal as written. As one
> > > > > commenter asked, "what if my kids want an IPv6 network to play with
> > > > > in their garage?" Well, we should find some way to accomodate that
> > > > > which doesn't require service providers worldwide to spend tens of
> > > > > thousands of dollars upgrading routers to hold the routes.
> > > >
> > > > Exactly. There's really no reason I should bear the cost of
> > > > carrying your route because your kids want to learn about IPv6. I
> > > > wholeheartedly want to support learning about IPv6, esp. for the next
> > > > generation of network operators, but doing so in a way that taxes
> > > > third party network hardware, for no reason, is not the way to do it.
> > > >
> > > > --
> > > > Garry Dolley
> > >
> > >
> > > i'm sorry - this smacks of shear laziness.
> > > trying to get ARIN to manage your routing table
> > > is kind of like asking your mom to still do your
> > > laundry.
> > Researching every /32 to see if it is worthy of being in your
> > routing table is not practical. It is not a laziness issue.
> your telling me what is practical for me to do?
> very kind of you indeed.
I did not mean you personally.
> however i beg to disagree - i ensure that i have the
> specifics i know i want and then proxy aggregate most
> everything else. there are even a few prefixes for which
> i -never- want to see traffic from and I block them.
> i'd posit that most folks do something very similar.
This model doesn't work for me, and I would imagine most ISPs.
> > > no one is -forcing- you to accept any route whatsoever.
> > > your router, your choice.
> > Yes, but practically speaking, I can accept all /32's or none of
> > them, with a little wiggle room with filters.
> i can't tell you what is practical for you.
> > I can set up filters, sure, but anyone running a multi-homed network
> > with real customers, peers, and traffic knows that maintaining
> > filters that actually work well is almost a lesson in futility.
> routing has not been "fire and forget" for many years.
> if you don't pay attention, your going to get burned.
> > > do the thought experiment... how many /32s are there in
> > > the IPv6 universe? Got a router for that? Didn't think so.
> > >
> > > Folk are going to have to face the fact that they can't
> > > depend on their benevolent RIR to manage the potential size
> > > of the routing table anymore...
> > But what we can do is try to promote policy that doesn't give out
> > prefixes like they are going out of style. Every /32 prefix
> > assigned and announced takes up one more RIB slot for me and every
> > other ISP on the planet. So, I'm going to do what I can to save
> > that resource.
> borrowing your metaphor, I'd like to see one family of
> prefixes come into style... and if that means treating them
> like loss-leaders for a bit, then I guess that is what needs
> to happen. Then, we can start worrying about them going
> out of style.
> I'll reiterate again. Just becauase I get a /32 and announce
> it to my peers, is zero reason for you to hold it (at all,
> or for any length of time, or in perpetutity). Folks
> really need to wean themsleves of the dependency on an RIR
> to craft their routing policies.
I'm not asking for the RIR to craft routing policies, even if the
policy does help / hurt routing. I'm asking for a /32 not to be a
default assignment without any restrictions.
ARP Networks, Inc. | http://www.arpnetworks.com | (818) 206-0181
Data center, VPS, and IP Transit solutions
Member Los Angeles County REACT, Unit 336 | WQGK336
More information about the ARIN-PPML