[arin-ppml] 4.4 Micro Allocations and IXP requirements

[Michael Richardson]

    mcr>         If two parties decide to start an IXP, and get a switch, rather
    mcr> than just do private peering, it's really hard to get to three if two
    mcr> don't count.  Still, one party or the other *ought* to have a /28
    mcr> around, and renumbering for two parties isn't that hard.

    mcr>         I propose a compromise: three parties (a route server would
    mcr> count) for IPv4 micro-allocation,

    david>     I think I like this idea.


Martin Hannigan <hannigan at gmail.com> wrote:
    > It's interesting, but you're introducing a new barrier. Capital. Some
    > are already encumbered by bad advice and capital constraints. By upping
    > the number a digit you actually increase their likelihood and getting
    > an ROI on their capital.

    > Adding a route server is a good idea, but without the third, it's a
    > waste of capital IMHO and a new barrier.

1) if you are an IXP, you need that switch.
   If it's an 8-port Linksys you got at FutureShop, I don't care.
2) I'm saying, you can *count* the route server, not that you have to have one.
   Or, you can count the third party.
   (A route server can be built with a PIII + quagga)

--
]               Never tell me the odds!                 | ipv6 mesh networks [
]   Michael Richardson, Sandelman Software Works        | network architect  [
]     mcr at sandelman.ca  http://www.sandelman.ca/        |   ruby on rails    [