[11084] in Commercialization & Privatization of the Internet
Re: A New CIX Design
daemon@ATHENA.MIT.EDU (Karl Denninger)
Mon Mar 21 20:48:57 1994
From: karl@mcs.com (Karl Denninger)
To: stpeters@dawn.crd.ge.com
Date: Sat, 19 Mar 1994 19:59:09 -0600 (CST)
Cc: karl@mcs.com, com-priv@psi.com
In-Reply-To: <9403200042.AA02632@swan-song.> from "Dick St.Peters" at Mar 19, 94 07:42:09 pm
> >From karl@mcs.com Sat Mar 19 17:40:56 1994
>
> >> there is more than
> >> one way to run a CIX.
> >
> >So put it on the table.
>
> Well, actually I have before, but let's try again. The main thing I'd
> like to see is to make CIX routing be a service that those who connect
> to the CIX buy and resell to their customers.
>
> Some elaborations on how to make this work and be equitable:
>
> Have the CIX be non-profit, so total CIX revenue is a reasonable
> approximation to total actual CIX costs.
The CIX is already a registered non-profit, and by law cannot retain large
sums of money and/or pay it out for other than direct services provided
(like salaries, etc). If there is ever evidence of chicanery or fraud
here, the people responsible will go to <jail>. Abuse of non-profit status
is a serious crime in the US.
> Leave intact the provision that anybody can string a line to the CIX
> and buy routing directly from the CIX, and have the CIX charge to
> directly-connected networks (DCs) be on a per-routed-network basis.
> These prevent new small guys from being shut out.
The problem with this is how do you pay for the equipment for those direct
connections? Have you priced CISCOs and T1 CSU/DSUs? With a Cisco 7000
running well over $50,000 fully configured (in fact, its quite possible for
such a beast to run over $100,000!) how do you spread <this> cost out
equitably?
If its on a "per-routed-network" basis you're going to have a lot of people
registering Class "B" networks and splitting those up to their customers so
as to evade the "per network" charge. This won't work if that starts to
happen. And, given this model, it will happen. Its unavoidable. Just as
people today put their SLIP customers on their own internal network numbers
in order to insure CIX connectivity (since you can't tell what's going on
then).
There is a <profound> difference in cost and load imposed between a Class "B"
network with 10,000 hosts behind it and a Class "C" network which can have
at most 253 hosts behind it.
If you impose the charge based on the network class you make it
unaffordable for those who have older Class "B"s that they registered back
when they were plentiful, but are using only a small percentage of the
address space.
How do you propose to address this?
The only <other> way to do this is what is called traffic-based pricing,
which comes back to settlements. The problem with <that> is that it
destroys the FTP archive sites (and WWW sites, etc) overnight -- hook one
of those up and all of a sudden your "pass through" bill from your provider
goes up by a factor of 10! The snowball effect of this means that you'll
start to see something like this:
ftp nice.archive.sit.com
Login: anonymous
Password: karl@mcs.net
VISA: xxxx xxx xxx xxxx
Logged in; VISA valid, transfers are charged to your card at $1.00/MB
Ftp>
Blech!
> The DCs can pass that charge on to their customers on any basis they
> and their customers find acceptable ... in other words, let the market
> decide.
Which they can do right now, if they so choose.
> Well, Chicago makes sense geographically, but I doubt you could get the
> big guys to come. I think it makes more sense to get the current rules
> changed.
The existing contracts and agreements run for two years. Anything which
dilutes the value we paid (and others paid) is going to be seen VERY dimly
by the present membership, and in fact could be seen as a breach of contract
by the current members. I know that I would view <very> dimly any attempt
to "force" the CIX to change its pricing model.
Then again, I don't happen to think its broken. It might not be optimum,
but it is the least-able-to-cheat model I can come up with, and the cost is
<reasonable>. Cheap? No. Reasonable for value received? Yes.
> Since I've never designed a CIX before, there are probably things I haven't
> thought of, but I think this meets your conditions and some of my own. It's
> break-even sort of by definition: add up all the costs, toss in something
> for growth, divide by the numbers of networks, and send out the bills.
You wouldn't like that result. Again, see above for why. The big folks
are going to cheat. No question about it. Its just too darn easy to do,
and in fact you can do so without any repercussion at all, since that's
all in the contract.
> To me, it looks equitable to big guys, small guys, old guys, new guys,
> direct-connects, and those who buy transit and routing from a DC. At
> worst a new small provider who can't get anyone to sell him transport
> and routing has to string a line to the CIX ... which is what he has to
> do now, but he won't be adding many networks to the pot, so his entry
> fee is small.
No -- he has to buy a card, and a CSU/DSU for the CIX end. This is close
to $5,000 at present prices. The line installation is $2,000, and about
$3k monthly in line charges for those outside of the LATA where the CIX is.
This is going to be MUCH more expensive for the average "small guy" than
the present model.
If you can't solve the "I'll split up my network number to evade the
per-net charge" problem (and you can't) you're really and truly screwed
with such a model.
> He can share the cost of that line with other small fry
> by selling transport and routing to them. I don't think this will be
> necessary though; if there's a market selling transport to the CIX,
> somebody will fill that niche. I even think I know who - or one of the
> likely several who's.
Perhaps. But perhaps not. You can't force anyone to sell transit. There
are only two "someone's" who will do it right now (at reasonable cost) that
I am aware of -- Sprint and ANS. Neither of them is "cheap" by any
definition you wish to apply, and both have strong incentives to keep their
prices relatively high -- like you competing with them. SPRINT is the most
like a phone company in this regard, which I applaud.
So you're shopping lines again for direct links. Get out your checkbook;
you're going to need it.
> Now where did I put that flame retardant ... ?
>
> --
> Dick St.Peters, Gatekeeper, The Pearly Gateway; currently at:
> GE Corporate R&D, Schenectady, NY stpeters@dawn.crd.ge.com
A good idea, but I don't think it would fly. IF there are enough providers
interested in this (like more than two or three) I'll run the numbers and
come up with some preliminary ideas on a model of this kind for a midwest
interchange point.
I think you'll find that it is, on the level, more expensive to do this
than it is to connect to the existing CIX under the existing terms. As you
grow you're going to find that this is <massively> more expensive.
Look at the present model again:
Let's say, for example, you have 10 people who want to corroborate on a CIX
connection. That's $10,000/10, or $833/year/each. That's a <$69> uptick
in cost monthly for routing to THIRTY other network providers and their
customers. Given that a 56kbps line cost is somewhere around $150 a month
for a LATA-based DS0, you're still talking something that a heavy voice
user would consider "affordable". And look at what you get for that
expenditure!
--
--
Karl Denninger (karl@MCS.COM) | MCSNet - Full Internet Connectivity (shell,
Modem: [+1 312 248-0900] | PPP, SLIP and more) in Chicago and 'burbs.
Voice/FAX: [+1 312 248-8649] | Email "info@mcs.com". MCSNet is a CIX member.
