[ppml] ARIN member in good standing?

Howard, W. Lee Lee.Howard at stanleyassociates.com
Thu Oct 5 17:49:07 EDT 2006

> -----Original Message-----
> From: Peter Sherbin [mailto:pesherb at yahoo.com] 
> Sent: Tuesday, October 03, 2006 12:06 AM
> To: Howard, W. Lee; ppml at arin.net
> Subject: Re: [ppml] ARIN member in good standing?
> > Can you describe the data capture mechanism?  I mentioned the
> > billing system that would be required.  Even if you discard
> > payload and only keep source and destination addresses and packet
> > size, you're talking about increasing network load by a 
> > significant amount (30%?).  Billing data for an OC3 could be
> > 16TB per month.  Multiply hundreds of circuits times a five-year 
> > record-retention policy, and I get a 97PB database.
> Assumption: need to count sent volumes only. A sending 
> network or a node (SN) is
> identified by its source address (SA). SN is connected to 
> provider's distribution
> router (DR). DR counts the length of all packets sent from SA 
> and reports the total
> number of bits to a database server. 

Yes, that's the math I was doing.  I had inferred from 
something you said earlier that you might charge different
rates per destination.  Even if not, if you have a 128 bit 
address field and a 16 bit length field, at a million packets
per second, it's still a couple of terabytes per month.   
That's a good-sized database, and that's just one OC3.

I think you suggest having a database server at every hub,
gathering data for all circuits there, and transmitting
summaries hourly to a central server, followed by a purge.
You're still tracking millions of packets per second, 
hopefully without introducing latency, and rack space isn't

> > I don't understand "unpaid volumes."  You bill your customers.
> In some cases volume based billing is announced but it may 
> not necessarily be
> enforced. Billing works OK for burstable circuits. A regular 
> T1 would be provisioned
> to accommodate e.g. about 0.5Mbps at a busy hour. Three T1s 
> on a single DR may have
> different usage patterns. E.g. one transmits at 1.2Mbps 
> during 3hrs in the morning.
> Second sends at 1.1Mbps at night for 6hrs. The third sends at 
> 0.3Mbps 24hrs daily
> with occasional bursts up to 1.5Mbps. Each pays $1,200 but 
> over the month #1 have
> sent 49GB, #2 = 89GB, #3 = 110GB. Sometimes providers would 
> go after "abusers" who
> exceed average usage tens or hundreds times. "Abusers" can be 
> looked at as an
> opportunity assuming they have a need to move data and are 
> willing to pay for it.

Sorry for leading the topic this way.  How and what ARIN 
members charge for services is completely off-topic.  If
your proposal requires Internet providers to bill in a certain
way, we don't have the authority to adopt it.  
> > You keep saying "proper," as if to say that there is an established
> > right way to do things.
> "Proper" means following the basic logic of a process.

I don't think that's what it means.

>  E.g. in a world of a regular
> mail or cargo shipments a sender pays for moving a certain 
> number of molecules over
> distance. Energy wise how does moving electrons fundamentally 
> differ from moving
> molecules? If electrons are too many to count let's define a 
> countable network unit representing a fair amount of 
> electrons.

I have it on good authority that the Internet isn't trucks,
it's tubes.  

The fundamental difference is that there's no Internet
oligopoly.  In shipping, you have a single carrier end-to-end.
In telecom, you may have two or three carriers, but they look
the same, because of heredity and regulation.

Another difference: we don't push electrons.  Individual
electrons don't move very far at all.  Data have no mass.

The basic issue at hand, though, is why the trucking model
would be substantially better than the status quo.  

> > I am not familiar with telecommunications licensing, but I do not
> > have the impression that licenses are available to anyone.  My
> > impression is that the "certain criteria" are high.  Does each
> > coffee shop and private interconnect have to get a license?
> The process is straightforward, as long as you meet all 
> "high" criteria you will get
> the license. If network services are defined and regulated 
> then anyone who falls
> under that definition is a subject to a license.

You say it's straightforward, but what are those criteria?
Does that mean that coffee shops need licenses?  If I set up
private interconnects between my enterprise network and four
of my vendors, do I need a license?

> > So does this mean a new kind of governmental licensing agency, 
> > which works closely with the government tax collection agency, 
> > and the two agencies direct ARIN?
> It is a matter of coordination between already established 
> agencies, e.g. Tax and
> FCC. Agencies feed ARIN with the number of subscribers by 
> provider as well as
> suggested fees. ARIN autonomously performs address allocation 
> practice. ARIN
> collects fees from all holders of licenses for network services. 

How strong are the suggestions?

> > I'm unclear on the taxpayer-node relationship.  Each taxpaying
> > individual with a tax ID gets an allocation?  What size?
> Taxpayers are not homogeneous and need segmentation. How many 
> addresses is enough
> for a single individual? Assuming that at some point we will 
> start fixing genes and
> that human DNA consists of about 3 billion base pairs will 
> /96 be enough? For the
> start we may want to reserve more space and start 
> experimenting with less.

Do you have a proposal?

> > Each taxpaying organization with a tax ID get an allocation?
> > What size?  
> Depending on what they can justify. Current ARIN approach 
> works well here.

I think your proposal is to have the tax and licensing 
organizations provide ARIN a subscriber count.  ARIN needs
more information than that.  Do we ask the organization for
more information, or train the tax and licensing agenices to
collect network diagrams and virtual host counts?

> > I was talking about IP address allocation.  Say my company has 
> > offices in 14 states and two provinces, with leased lines between 
> > them, and three Internet connections.  Do I get three assignments 
> > from my carriers, or 14?  Or since I have leased lines, do I get a 
> > telecom license?
> You will get your PI assignment directly from ARIN in the 
> amount you can justify.
> Whichever provider(s) serve your Internet connection(s) they 
> will report it to ARIN.
> You will get a telecom license if you meet all of the 
> criteria for it in your region
> and if you want to be engaged in a telecom business and if a 
> contract with your
> provider permits you to sub-lease lines.

I still don't quite understand the threshold where a telecom
license is required.  The only telecom service I want to 
provide is to employees of my company.  Do I need a license
in your model?  Do I meet the criteria?  If the criteria 
include maintaining a billing database and providing wiretap
facilities, then I don't.  Would that mean my private WAN is

[redacted:  addresses for me, my wife, and her home business]

So we could easily have three non-aggregatable prefixes at
our house.  Every organization and every individual would
have a non-aggregatable prefix.  Do you foresee any routing
scalability problems with that?

> > What's a top level licensee?
> E.g. FCC in the US or CRTC in Canada.

Fees would be variable between countries, and set by the
government, not by ARIN and its members.

You propose significant regulation of the Internet and what
look to me like significant additional costs for Internet
access providers, both of which may (depending on how well
I understand your proposal) reduce the number of providers
and the flexibility of enterprises to bypass them.  Who
benefits from this?

Or, put another way, I don't understand the problem we're
trying to solve. 



More information about the ARIN-PPML mailing list