Aaron France writes:
Unless we can implement a way to bill eachother for traffic passed
through eachothers pipes (reciprical billing, like telco's), they have
no right to assume that they can bill us because their network is bigger.
Theoretical question: What if ISP-X had only one dedicated connection behind it, and subsequently was 1/100,000 the size of NSP-Y. ISP-X's only connection was a single content provider: Yahoo.
How do we resolve the settlement issues in this model? Why should the ISP-X pay NSP-Y to deliver valuable content to its consumers (dialup and dedicated)? Does the possession of more CIDR blocks dictate value? In today's market (of significant price competition, where UUNET is not terribly competitive), UUNET would probably experience substantial erosion of its client base. Wall Street and shareholders should pay particular attention to UUNET's follow-through of this policy as it appears to have the potential to impact its market share (and subsequently its revenues).
Traditional broadcasting models work in reverse of the UUNET model, but again, we're focusing on the wrong layer for settlement; i.e. the value is in the content, not the CIDR.
Having UUNET charge peering ISPs for application content on their network is like having Hughes charge Cox, TCI, and other CATV networks for receiving CNN's broadcast over the satellite, while also charging CNN to carry their signal. It is my understanding that the settlement is between the CATV network and CNN, and Hughes is only transit paid for by CNN out of this settlement revenue.
As diagrammed (with Internet / CableTV models referenced), the settlement model should look like this:
consumer pays local access provider (ISP / CATV)
local access provider receives content from transit provider (NSP / Satellite)
transit provider is paid by content provider ("webcaster" / Cable network channel)
content provider charges local access provider
Looking at a few other industries, there are significant parallels. With postal service, the producer of the content usually pays the transit provider to deliver the content to the recipient. I.e. Compaq pays UPS to ship a new computer to me. I pay Compaq for the content. Of course, items can come "postage due," but not typically when the sender has also paid postage. Does this infer that content producers in the Internet are not paying 100% of the delivery expense? Considering there is no charge between UUNET subscribers outside of the fee each pays to connect (i.e. the process of sending packets to another UUNET subscriber does not create additional charges outside of the base fee), this does not seem to be the case.
What then is the origin of the peer point charge? The expense of delivering the traffic to the exchange for consumption by another external network? Wasn't this paid for by the content producer (i.e. what is my $3,000 / month going for, if not for transit to exchange points? Local loops only cost $340). Again, it appears that freight has been paid for already in the UUNET model. It appears that the primary motivation for UUNET's move can only be to eliminate competitive pressures from mid-sized ISPs by restricting financially viable peering to a handful of providers, subsequently increasing network service prices.
Putting our company into the above model, Intellitek (a content provider) is not interested in charging local access providers for access to our products at the current time. Marketshare is of significant value, and we value the unrestrained access to our services consumers of the respective local access providers. ISPs represent a significant consumer base, and unrestrained access by these consumers to our content services is key to the success of our application.
However, if our transit provider (UUNET) suddenly impacts our ability to 1) set our delivery policy (presently at zero charge per consumer) and 2) limits the extent of market reach in its attempt to profit from our content, I would imagine that we would not continue the use of such a transit provider, with one exception.
If we balance the settlements by charging the transit provider, we subsequently create a situation where charges are being passed on. I.e. UUNET becomes a wholesaler of Intellitek application services. UUNET benefits from content consumed from our application service through peer points. Content is subsequently transfered to UUNET's domain through a distributor arrangement for delivery to ISP "retail outlets."
This may work in an environment where peering is not restricted on a casual basis, and multiple transit arrangements are available. I cannot imagine a Compaq or Digital limiting distribution of its product through only one wholesaler -- and content providers equally would not support such a relationship with its transit provider.
I expect that if UUNET continues its course, content providers on its network will either migrate to other networks to ensure delivery of their content to their market without restriction, or impose settlements on UUNET that balance the economic model and open their networks multiple transit providers.