craig@NNSC.NSF.NET (Craig Partridge) (04/28/88)
Perhaps I'm covering ground already mentioned somewhere (if so,
set me straight) but I wonder if charging by packet is the right
mechanism.
Consider charging people entirely by the capacity of their link
to the Internet. I.e., if you have a 56Kbit connection you pay a certain
sum, if you have a 9.6Kbit connection you pay much less and if you
have a T1 connection you pay much more.
These cost are set in such a way to cover the infrastructure costs
(e.g. the cost of the backbone networks).
The logic behind this scheme is that your line speed tells us roughly
how much traffic you can introduce into the network (as either a sender
or receiver), and if the costing is done right, you have an incentive
to pay for only as much bandwidth as you need.
Similar schemes are in use in other fields. Pardon the choice of
example, but in cities where you pay for garbage collection, an oft
used payment scheme is that you (even residential customers) rent
approved garbage dumpsters from the city -- and the city only picks
up trash in approved dumpsters. The result is that you buy as much
dumpster capacity as you need to get rid of your trash (and you have
some incentive to minimize your trash production).
Of course this scheme can be rigged slightly so that lucky folks
in high network density areas are paying more than their fair share of
the infrastructure costs so folks in low density areas can get networking
too.
CraigAI.CLIVE@MCC.COM (Clive Dawson) (04/28/88)
Another scheme which might come in useful, and which is often used by city utilities, is to pick a period of time where usage is monitored to obtain a figure for "average" usage. Then a charge is based on this figure. There are no meters on our sewage lines which can be used to bill for wastewater. Instead, the city bases its wastewater charges by sampling potable water usage for a month or two. (This usually is done in the winter time to avoid the lawn watering factor, etc.) In this case, the network administrators might wish to compute an "average" packet count for each host by monitoring for some period of time. (Such times should obviously be well-kept secrets so that hosts don't influence the result by changing their behavior.) Network charges could then be assessed based on these results. Just figured I had to get MY two cents in...! Clive Dawson -------
ittai@VX2.GBA.NYU.EDU (Ittai Hershman) (04/28/88)
I have been sitting on the sidelines watching the discussion on network costing/charging. I am afraid that some people have missed the forest for the trees. Craig Partridge has put us back on the right track. The Internet is a tool to increase R&D and productivity. Because this increase is a result of sharing among competitors (regardless of whether these are profit-making or non-profit institutions, sharing of information results in a loss of competitive edge). Yet it promotes a common good and the rules-of-the-game are changed by this sharing: standards are formed, technology is advanced, and competition is spurred. This, in my opinion, is precisely why the Internet is an "infrastructure issue" as are interstate highways; like interstate highways, the cost of the network infrastructure -- the PSNs and the trunks which interconnect them -- should be federally funded from general revenues. I suspect that the Internet infrastructure operating and capital costs can be budgeted with a high degree of certainty. Furthermore, the States have gotten involved, and we now have a bunch of regional networks with State funding as well. So the model is one of a federally funded national infrastructure which serves two subordinate layers: federal institutions, and regional subnets. The regional networks, in turn have subordinate institutions. And finally, the institutions are responsible for their own internal networking. All of this, I think, conforms to the spirit of FCCSET. The pragmatic problem, of course, is that it is (unfortunately) unlikely that the Congress and the State Legislatures will value the Internet as we (who are biased consumers of it) do. And so, in this imperfect world, we must find a way to raise some of the money ourselves. And since many of the consumers of this resource are cash poor, we struggle to find an equitable method of cost distribution which does not negate the raison-d'etre. I agree with Craig. Charging by "bandwidth" is both equitable and reasonable. Look at the rousing sucess of BITnet -- it permitted virtually all institutions of higher education an affordable but minimal level of service. A lifeline service. Mail. To get more, you pay more. To further raise revenues, I think the Internet should allow/encourage involvement of commercial companies. Aside from the computer manufacturers and software houses, there are many commercial companies involved in (non-defense) research and would benefit from Internet access. For example, the financial institutions in the Wall Street area hire PHDs (and faculty) galore and have more advanced computer technology than most CS departments. Many would be interested, and certainly could be "sold" on joining. They would be charged higher fees, since their return to the Internet will be less direct. The fees could be used to subsidize the non-profits. (Note that we would not be selling communications media, rather the data, so I think this can be justified legally). Lastly, I suspect that the most expensive (read: inefficient) part of the Internet at present is not the traffic per se. Rather, that the redundency is at the site level rather than at the infrastructure level. NYU, for example, has three Internet connections: MILnet, JVNCnet, and NYSERnet. Wouldn't the resources required for all of this (computers, gateways, leased lines) be better spent on the PSN/trunk level?!? Reorganizing the whole mess will probably save enough money to make some needed capital improvements. On a final note, as a member of the University-wide network planning committee at NYU, I see that we are a microcosm of the Internet. We are a huge institution, spanning unwieldy amounts of real estate, with decentralized control. We have "rich schools" and "poor schools" and the rich subsidize the poor. We have not found a good method of charging for network services, and I suspect we never will. You can't quantify the loss of an infrastructure service -- only what it costs a part to run, and what it costs to replace that part if it fails. The value of the whole is immeasurable. Apologies for the length... -Ittai