The following text is copyright 1996 by Network World, permission is hearby given for reproduction, as long as attribution is given and this notice is included.

The "S "word

Some words seem to evoke magical healing powers. Just say them and somehow the path is clear and the future bright, the frail made strong and broken made whole. The networking business has had its share of magic words over the years. In succession we have seen people invoke words like ISDN, OSI, and ATM (that is if you can call such acronyms "words") as beacons to one future or another. Each and every one came with the promise of the glory of easy and dependable networking to meet our every need. Current favorites include TCP/IP along with Inter- and intra- net.

A new word with the claimed ability to fix the Internet (whatever that means) is starting to be chanted in various circles. But at the same time this word strikes fear into the hearts of some. The "S" word; settlements--payments between Internet service providers (ISPs) based on the traffic flow between the ISPs.

Some people argue that if we had settlements between ISPs as we do among telephone companies all of the problems of the Internet will be solved almost overnight. If one ISP needs to pay a second ISP when it sends traffic to it then this will mean that the 1st ISP will insist that the 2nd ISP be reliable. Others fear that the establishment of a requirement for ISPs to pay each other to exchange traffic will mean a fundamental change in the way people pay for their Internet service. The $19.95/month flat rate access would go the way of the 300 baud modem. There are problems with both of these positions and with some of the basic assumptions leading to the thought that traffic-based settlements might help fix some known problems.

The traffic flow between the largest ISPs turns out to be roughly symmetric. So if traffic-based settlements were to be established whatever checks that were sent between them would balance out, in fact if each ISP were to send a check based on the sent vs. received balance of traffic, the checks would be for very small amounts and over a few months generally cancel each other out. So, for the big providers, the cost of the establishment and operation of a traffic-based settlements process would far outweigh the value of the return from the use of such a system. Somehow I do not see how getting the big ISPs to exchange small checks will do much for Internet reliability.

There is little issue of settlements for most of the small ISPs since they are customers and not peers of larger ISPs and are treated the same way that any customers of that ISP are treated.

There are different factors to be considered when talking about the relationship between medium sized ISPs and the large ISPs at the exchange points. In these cases the traffic balance may well be uneven and having the smaller ISP pay the larger for the service of peering may give the larger ISP some incentive to provide a better service but it might be a good idea to be sure that the cure is not worse than the disease. In any case it is clear that just requiring traffic-based settlements will not fix whatever problems you currently find with the Internet.

disclaimer: The above must be my opinions since settlements at Harvard tend to be rather more one sided rather than by-lateral; if Harvard agrees then you can pay a bunch of money to attend classes (the learning part is up to you).