Internic address allocation policy

Paul A Vixie paul at vix.com
Sun Mar 19 18:16:21 UTC 1995


i'm not sure i agree with karl here.

>But the new ISP has no basis to request these.  However, the CURRENT NIC is
>declining requests for *256* Class "C" addresses!
>
>That is, I was turned down for a Class "B" equivalent in a CIDR block.  That
>is ludicrous.  Any reasonable regional ISP, large or small, is going to go
>through that in a year.  And if you enforce a 75% usage requirement to get
>more, then you've got something workable.

this begs a question (which karl will proceed to try to answer), which is:
how does the NIC know whether a new ISP is worthy of even 256 class c's?
in karl's terminology, how do they know whether an ISP is "reasonable"?
sure, "anybody knows," but i don't think this is an area where we can just
depend on the NIC to decide based on their personal knowledge of the ISP
or the industry.  it's not that i don't trust them -- i do, actually.  but
the potential for liability on the NIC's part if they don't have -- and
follow! -- objective allocation standards is too high to ask them to bear.

the 75% usage problem is rather hard to verify, as well.  to get a negative
answer you'd have to know during the host count that the network being
examined was at that moment routable.  or you'd have to depend on the ISP
to do its own host counts, which opens up horrendous fraud potential.

>I would define "usage" as "has a routing entry active on the net". 

i've got 9 bits of address space in the core routing tables right now.  if
i tell the NIC i'm planning to enter the ISP business, does my existing
route entitle me to a block of 16 more bits?  if not, why not?  here's
the crux of the problem:

>Note that this does have an honesty component, as, for example, we have
>part-time networks connected via dial-up which are only routed when active.
>But trust me, we have issued what we asked for -- and that space IS being
>actively used by real, live, paying customers.

i know that.  and if anybody asked me whether Net99 or MCSnet needed 16 bits
per allocation, i would say "hell yes!" since i know you guys (a) know the
meaning of what you're asking for, and (b) will use it wisely and honestly.
i cannot depend on either (a) or (b) for the average new ISP-wannabe who
has sold their video rental store and wants to reinvest the proceeds in the
Internet 'cuz they saw Al Gore on Tee Vee and though they don't know what the
Internet is, they know they gotta have some, and isn't that book by Canter
and Seigel just the greatest thing you ever saw?

hell no, i won't go.  along that path lies chaos.  what we need is some kind
of "ISP Council", with core membership determined by some combination of
customer votes and hard dollars (which keeps out ignorant newcomers and large
providers whose customers don't like them but don't hate them enough to switch)
and secondary membership by invitation of a majority of the core members (thus
making room for people like kent england or even your humble author).  this
council would meet electronically every so often to ratify allocations which
are outside a conservative upper limit, and meet physically every so often to
argue about whether the underlying allocation scheme or current limits need
to be changed.

nothing i've seen on com-priv or cix-members or nanog or rfc1466bis has yet
addressed the fundamental problem of using government money (or allowing any
government to affect the policy) for internet resource allocation.  we, the
users (and the greater "we", the providers) have to take this over.  we've
got the dollars at risk and we've got the customers to satisfy -- and those
two things are the primary components of networking.



More information about the NANOG mailing list