[arin-ppml] Rationale for /22

[William Herrin]

On Fri, Jul 31, 2009 at 3:31 PM, Leo Bicknell<bicknell at ufp.org> wrote:
> I suspect most of the folks popping up and saying they had 6-10
> different /24's in the global table before getting an ARIN allocation
> either did it prior to us relaxing the limits (at one time it was
> a /20), or didn't want to deal with ARIN for some reason.  Today I
> see no reason why a someone should need more than 2-3 blocks before
> qualifing for an ARIN PI assignment.

Hi Leo,

>From their descriptions most of them were small ISPs rather than end
users, where the standard is still /20.  But I agree with you that we
should take cautious steps forward.

What is a cautious step forward? /21 for ISPs and /23 for end users?
Perhaps. On the other hand, we have described some issues during this
discussion that could be exacerbated by just blanketly lowering the
bar.

On the flip side, the latest 2U DL380 servers from HP have 8 CPU
cores, 144 gigs of RAM and 16 300gig hard disks. How many of these do
you need at 1 IP address each before you've created an Internet
service of sufficient value to merit multihoming?

Would it be reasonable to consider bridging the gap between /20 and
/24 with an entirely different policy? Something that more directly
attacks the "whose routes and how many can we afford to carry" problem
instead of the sideways "250 machines now" requirement that a /22
imposes?

Perhaps tackling the problem from a different direction could get us
to /24 with a step that's more cautious and more conservative than
just lowering the limit to from /22 to /23.

Regards,
Bill Herrin



-- 
William D. Herrin ................ herrin at dirtside.com  bill at herrin.us
3005 Crane Dr. ...................... Web: <http://bill.herrin.us/>
Falls Church, VA 22042-3004