Efficient (Dense) Use of Address Blocks

Karl wrote:
> > My request was for 256 Class "C"s. We have consumed almost
> > exactly that in a year.

And Alan added:
> I did the same. [...] Interpath has
> not *QUITE* used up the block, but pretty darn close

Can either of you provide estimates of how densely these blocks
are populated? What percentage of the individual host addresses
are currently used by functioning hosts (or router interfaces)?

In some cases, 99%. Really. We hand out single addresses and route them in
many cases, and those addresses get used until they're gone -- all of them,
except for all zeros and all ones.

Some blocks are less densely populated. If I subnet to 26 bits then 1/2 of
that Class "C" is wasted, which I can't control due to the technical
requirements of the network involved.

Perhaps you're both saying you re-assigned use of these address
blocks quickly, rather than saying you populated them densely
with actual hosts?

Uh, no. We populated them as densely as reasonably possible given the
requirements of our customers and our own internal engineering allows.
We have spent a hell of a lot of money, and made a number of purchasing
decisions, based on whether or not certain products could or could not do
VLSMs and the like, and whether they could handle host routes. All in the
name of not wasting address space.

One could imagine ISPs (not Karl or Alan, surely!) who might
re-assign blocks of class C (octet-wide) networks when a single
one would suffice. Or they might assign octet-wide networks when
quartet-wide (nibble) networks would suffice; or even quartet-wide
networks when triplet-wide networks would suffice.

Until there is a solution for in-addr.arpa quartet-wide or triplet-wide
network assignments to individual customers who want to run their own DNS
(and many do for VERY good reasons) is damn close to impossible. This is a
technical problem which has not been solved in the DNS system for a long
time, and it stinks, but its a reality of life right now.

If and when it is fixed we'll adjust our allocation policies to match the
new reality of the world. Until then we have to live with the reality as it

You want to know what would, more than anything else, help the address space
issue? Solving the in-addr.arpa problem. Solve that one and I can reduce a
number of my customer allocations, and so can all the other providers. That
could, I would estimate, cut the growth rate of address requests at least in

I believe ISPs who receive large blocks have a responsibility to
make sure they (and their customers) use them efficiently. How
efficiently, you ask? If you're not getting more than one U.S.
dollar of value per host address per year, you're not using it
efficiently in my book!
Sean Shapira sds@jazzie.com +1 206 443 2028
<a href="http://www.jazzie.com/sds/">Sean's Home Page</a>
               Serving the Net since 1990.

We're getting significantly more than one US dollar of revenue per host
address per year from our allocations. In much of our address space we're
exceeding that metric by more than two orders of magnitude.

with ciscos ;you can use subnets all ones and all zeroes.

So in fact you can use 2 subnmets in one class C with the 'bad' mask .
Maybe the other vendors should do that too, it is a question of customer
Explain why you need it and why it is possible without creating chaos,
and the vendor will certainly do it (Cisco did for al lones, e.g.)