[1894] in Commercialization & Privatization of the Internet
Understanding Combits
daemon@ATHENA.MIT.EDU (Jordan Becker)
Mon Jan 6 15:16:28 1992
Date: Mon, 6 Jan 92 15:11:58 EST
From: Jordan Becker <becker@ans.net>
To: cook@tmn.com
Cc: com-priv@psi.com
> <<MESSAGE from>> Gordon Cook 03-JAN-92 23:34
> cook@tmn
> Gordon Cook speaking: So what is measured then? Is it the amount of
> combits going into SURAnet as compared to the amount of combits coming OUT
> of SURAnet? If SURANET puts out more combits onto the back bone than it
> takes in, it OWES ANS money. If ANS shoves more combits down SURAnets
> throat than Suranet shoves back, ANS owes SURANET money?
A COMBIT is a directionless unit of measure that quantifies the flow of
traffic between two unique network numbers. COMBITs are collected at a
gateway between two service providers and may be defined as:
COMBITs = [packets_in + packets_out]*300 + [bytes_in + bytes_out]
How this formula was derived and is measured is documented in an appendix of
the ANS Plan for Commercial Services and may be beyond the scope of this
discussion. The point is that COMBITs would be measured at an ANS-SURANet
gateway for each SURANet subscriber network number that participates as a
source or destination network, in a traffic exchange with the ANS network.
The COMBIT total for any specific SURANet subscriber network would be counted
as CO if the subscriber network was declared by SURANet as commercial, and RE
otherwise. Therefore, assuming that SURANet does not declare any of their
subscriber networks to be commercial, then they would have 100% R&E COMBITs in
their traffic exchanges with the ANS network, and as an NSF sponsored midlevel
they would pay nothing.
> Can I ask the question differently? Can SURANET sign the connectivity
> agreement with no obligation just so long as it doesn't have a customer
> that declares itself commercial?
Yes.
> And while the explanation we see here appears to focus on a new commercial
> user's impact, what the ANS agreement focus on is taking the PAST traffic
> patterns of the mid-level and figuring on the basis of THAT what the next
> year's bill is.
Actually this could be forecasted on a quarter-by-quarter basis so that a
midlevel can predict how to charge the institutions they declare as commercial
in order to recover the ANS incremental CO costs from them.
> For what? For an infrastructure pool that Al Weis says
> is now worth $25,000? Lets say that ANS gets 100 NEW customers on its net
> in the next year. My impression is that this would make the
> imfrastructure pool grow to about $500,000?? In view of the fact that the
> NSF is spending about $16 MILLION dollars a year on support for the
> mid-levels totally apart from the backbone, how far does $500,000 go?
It depends on how many subscriber networks declare themselves as commercial,
and how much incremental costs are incurred because they are declared
commercial versus R&E. Remember that these are purely incremental charges due
to the change in classification of a subscriber network from RE->CO.
These charges do not include all of the fixed infrastructure costs that are
recovered from direct ANS attached service subscribers. These funds are used
to operate and enhance the ANS network infrastructure, which benefits all
attached networks.
> And
> why should commercial information providers sign up with ANS and find out
> that they are unable to reach mid-levels that haven't signed the
> connectivity agreements when they could sign with a mid-level, or with PSI
> or Uunet and be able to reach ALL the mid-levels for less money?
They could also sign up with ANS under the R&E AUP and reach all of the
midlevels for less money. Some subscribers will want to pay extra for a
commercial restriction free AUP. Those that do not want to pay extra have
other options, some of which you have enumerated. It is simply our intent to
offer something that did not exist before, a non-ambiguous way for an ANS or
midlevel service subscriber network to declare itself to be free of commercial
use restrictions.
Jordan
