A Practical Navigator for the Internet Economy



We interview Vijay Kumar, the lead designer of Lucent's Packet Star terabit, IP-in-silicon routers. They have thrown out the traditional architecture methodology of having a CPU do software lookups on every incoming packet in order to decide how to treat that packet both for routing and QoS purposes. To avoid both the processor to memory bandwidth choke hold they transfer packet processing functions to hardware. Here they gained such an opportunity by using programmable gate arrays running at about 30 MHz to build a mechanism that by making binary decisions can sort incoming packets into as many as 64,000 queues.

These queues can, in turn, be set up in such a way as to permit the application to them of Quality of Service criteria. Why? Because this enables a separate queue for every flow of every user. The flows will be very bursty and, as such, the packet forwarding engine part of the router that we have described is very good at multiplexing. These silicon packet sorting engines are positioned both at input and at output ports.

For example, all packets originating from a certain location of company "a" and going to a certain range of IP addresses at company "b" get treated in a certain fashion. This grouping is necessary because, for example, this location of company "a" has several thousand employees. Entering a separate rule for each employee, would be too complex.

Another rule might say: all packets coming from this particular department at company "a" and going to a department at company "b" will be treated according to a prior agreement. As a result, when a packet comes in, there are overlapping rules that apply and they had to figure out which set of those rules is the best match. These kinds of things really complicated their lookup scheme. It was a breakthrough when they developed an algorithm to do this. T.V. Lakshman and Dimitrios Stiliadis published the algorithm in the last ACM Sigcomm. They also published an expository article in the May 1998 issue of IEEE Communications magazine.

They noted that the growth rate of IP-only nets is phenomenal and use of the legacy protocols is shrinking. So they like Avici, Juniper and Pluris designed an IP-only router. But they needed to differentiate themselves from the rest of competition by adding quality of Service to fast IP. Such an achievement would enable VPN applications where a national ISP could use the Packet Star queue assignment mechanisms to establish using IP only the equivalent of ATM permanent virtual circuits between customers.



We interview Henry Sinnreich, MCI's lead IP telephony designer and director of its IP telephony engineering effort. He gives his assessment of some of the key protocols and development work now underway. Mapping of IP networks to PSTN phone numbers will involve E.164 and SIP. E.164 is an ITU global addressing standard that basically gives the country code, area code, and user number in the local telephone switch. Unfortunately, it only explains how telephone networks should be numbered.

The Session Initiation Protocol SIP is useful in dealing with telephony based URLs. One can enter such a URL in a browser-like SIP client and type sip: +1-972-498-1223@my carrier.com. The IP network software deals with the SIP URL much in the same way as it deals with other web addresses, and this enables developers to re-use well understood, true and trusted code for HTTP.

Henry reserves a place of special condemnation for H.323 as a standard used in early internet telephony efforts. He points out that H.323 is an ITU video conferencing standard promoted by a consortium of video conferencing vendors called the International Multimedia Teleconferencing Consortium (IMTC). It has unfortunately and mistakenly been pushed also as a standard for IP telephony by people (video conferencing oriented vendors) who did not have to build and operate either telecom or IP networks. Actually the ITU H.323 document says that H.323 applies to packet networks, such as IP and IPX LANs, but it does not pretend to be a standard for voice over the Internet.

He points out that H.323 wastes large amounts of bandwidth and forces users into a management Babylon of dealing with multiple, on-going, complex systems using OSI based management software instead of IP based management software. It offers an interoperability dead end, in part by using neither telephony based NPAs nor Internet based DNS. It uses non standard security devices and is world wide web unaware.

On IPDC: IPDC is probably a well-designed protocol suite for the control of IP telephony gateways for the state of the art 6-9 months ago. It is also a fascinating example of the different cultures; bellheads and netheads at work. . . . For the Internet, it does not deal with such items as flexible addressing, IP security, variable delay, packet loss, etc. For this reason he believes IPDC to be a good design, not necessarily the best, for situations when all devices are inside a box, and possibly on a LAN, but not spread out across an IP network.

Henry concludes with a very fascinating discussion of a block diagram showing, on the one hand, the complexity of the hybridized IP and PSTN and, on the other hand: the simplicity of the future IP-over-glass network that goes from DWDM, to gigabit ethernet framing, to IP and then benefits from a fully IP native applications layer that, being all packets, does not discriminate between kinds of data. He explains why such a network runs at a fraction of the cost of the PSTN.


The ISP friendly national backbone, Priori Networks, open for only about a year, went bankrupt in August. We present a short post mortem discussion among ISPs of the lessons learned.



In a 28,000 word report we try to put perspective into the formation of the new IANA Corporation now known as ICANN. It is a process which very nearly has fallen completely apart at its end: marked by a Congressional inquiry targeted at Ira Magaziner and the Commerce Secretary, and by the tragic death of Jon Postel on October 15 and 16th respectively.

We have undertaken a total revision and up-to-the-minute expansion of material placed on our web site during the first ten days of October. We attempt to trace the dual events that took place during the summer which was marked, on the one hand, by the IFWP process and, on the other hand, by the conclusion of the Commerce Department's negotiations with NSI and a two year extension of the cooperative agreement. We try to place both processes in their global context marked, on the one hand, by American European tensions over the looming EC deadlines on American non compliance with EC privacy statutes and, on the other, hand EC dislike of both NSI's "monopoly" and its tension over the US technology leadership in the globalization of the Internet. The ICANN formation cuts to the heart of how the technology of the Internet will reshape, first of all, global telecommunications and secondly world trade, the majority of which is expected to be done on the net within the next five years.

What appears to us to have been a sellout by Ira Magaziner to the interests of an inflexible Joe Sims, acting in a pro bono capacity, not for Jon, but for the IANA function, may only have been Ira Magaziner's having to adjust his policy to Sims' failure to deliver to Ira what Ira believed he had committed to. We document extensively how Sims worked with a small group of Internet old boys to choose a board and break out of his consensus agreement without Ira's prior knowledge or approval. We also document the possibility of conflicts of interest between Sims' law firm of Jones, Day, Reavis and Pogue and likelihood of having the open, accountable and transparent new IANA Corp. called for by the White Paper.

Two conversations with Ira on Monday October 19th have given us hope that Ira understands how forcefully he must act to ensure that Sims, who is now negotiating as the IANA principal , takes the consensus process mandate by the White Paper Seriously. A letter to be delivered today to Herb Schorr, the director of ISI, and signed by Magaziner will point out in detail how Sims will have to negotiate major changes in the bylaws of ICANN and related conditions for the corporation's operation before the Administration will approve ICANN's privatization of DNS and related IANNA functions.

==================== From our LONG article on the ICANN/IANA issues =====================

Until the bombshell of the Congressional inquiry hit on October 16 and the sad news of Jon's death arrived on the morning of the 17th, it looked like we had arrived at the finish line for a decision that will irrevocably shape the future of the Internet.

And now, after talking with various people including twice with Ira Magaziner on October 19, only the following is clear. Sources say that Magaziner believes that he has tried to run an open, fair, and honest process. He believes that the letter to be issued in the morning will enable the award of a cooperative agreement to ICANN that will be the result of a process by which ICANN adopts the most critical portions of the BWG and ORSC bylaws. He further expects to have a clause in the agreement that will ensure that the selection of the final board will truly be open. Finally he hopes that this will pave the way this weekend to the acceptance by all parties of the true consensus agreement that he says he has always been seeking.

However, they say that he was caught off guard by the events that flowed from his belief that, having said the private sector should create the entity, he could afford only to suggest to the participants that they should do certain things. He believed that he had their commitments and trusted them to live up to those commitments. But he found out only after carrying out the discussions aimed at finding the necessary consensus with the Boston Working Group and ORSC, among others, that many requests he had made to Joe Sims to negotiate with these groups and seek consensus were ignored by Joe. Neither group had ever received any contact from Joe. Also Ira is said to be very displeased the results of asking IANA and Joe Sims to have a list of board candidates out for public discussion by the second half of August. When this date was missed he then called loud and clear for the 23rd of September.

That although Ira expected draft five, and had demanded the removal of clause 4.1 (d) and (e), Ira did not know that it would no longer be a consensus draft between IANA and NSI. That he talked with both Jon Postel and Joe Sims on Friday September 25 and Tuesday the 29th. That the onset of Jon's final illness occurred a day or two after the conversation on the 29th. That not only did Sims fail to bring any names as nominees out before September 30 but, of course, that when they came out on October 2, the board was a fait accompli. That this was a major failure for the process that Ira had asked Simms to assure him that he would not allow to happen.

When today I asked why the government was giving the agreement to Joe Sims since Joe's client, Jon Postel was no longer alive, I was told that Sims has a letter of agreement for his legal pro bono services signed with Herb Schorr the Director of ISI (Jon Postel's home institution) on behalf of the "IANA function." That Sims was always the attorney for the IANA function and never functioned as Jon Postel's attorney. That ISI is now the client and will be the receiving entity for the cooperative agreement, if Simms negotiates in good faith. (We hope that ISI has provided Ira with a copy of its agreement properly dated. Some readers may remember that this is the same ISI who effectively with drew its legal umbrella from Jon in its letter of April 4, 1997.)

Several sources I then talked with found this to be shocking news. So I have made two calls to Sims' office to get his response. The first call left the message with the secretary. The second time I asked whether there was a response. No. Did you talk with him? Yes. So he got my message? I didn't say that he got your message, Mr. Cook. Well did he? Do you have anything else that you want me to tell him? Yes, that I would appreciate an answer. Goodbye. I made a call to Herb Schorr the director of ISI. Does ISI have a contract (letter of agreement) with Sims? Yes or no? My call was never returned.

This whole dodge and feint process is an example of what is wrong with the whole process the entrance of Joe Sim's. What we have is the appearance of a pro bono attorney where the circumstances of his relationship with Jon Postel were kept (one must assume by ISI and Sims) effectively hidden. I asked Sims earlier in the summer on several public lists when I pointed out that Sims was working without charge for Postel to reveal whether he had a letter of agreement with Jon. He did not answer. I am only sorry now that I did not press him for it.

I have heard some extraordinary assertions from the US government about Joe's lack of cooperation with the same US government. In this case it appears that Sims has arrogated to himself the primary responsibility for negotiating further changes to the agreement. (Another uncertainty is the question of who signs on for the fiscal responsibility involved in the transfer of government assets.)

I am very glad that Ira has made it very clear that the letter to be sent to Herb Schorr of ISI in the morning will explain that Joe's bylaws are deficient in a lengthy list of issues involving of political transparency, fiscal accountability and many other areas common to the complaints of both the BWG and ORSC. That this letter will make clear that, if the US government is to work with Schorr and ISI in handing over US assets to the ICANN corporation, Simms must agree to enter negotiation with ORSC and BWG. That these negotiations must be rapidly completed if Simms is to demonstrate that he is working this time in good faith with the government. That when the negotiations are complete both ORSC and BWG will be asked by the government whether or not they accept the bylaws that Simms puts forward. Only then will the government begin to negotiate the completion of the agreement with ISI and Sims.

Ira, it seems, has appeared to be operating in bad faith because things have gone differently than he had been led to believe they would by Joe Sims. What Ira said he expected to happen did not always happen, especially last month and this. We have had perhaps 15 conversations with him since November 1, 1997 and perhaps six of those since September 21, 1998. One thing that we find very striking is the unwavering consistency of the process that Ira has said that he has attempted to follow. Others have noted this consistency as well. We have said some harsh things about him recently. If he can bring off what he has outlined today that he intends to, the Internet will owe him a debt of gratitude.

For the unthinkable has come very close to happening. Ira can not permit the man who presided over such a non responsive process as the one we have seen to walk off with the first legal charter of the Internet that would have been essentially accountable to no one except the Attorney General of the state of California. Stop and think. Joe's firm will in all likelihood become corporate counsel to ICANN. The Technology Issues Practice (TIP) of Jones, Day could have been then in the singular position of having created a governing entity especially amenable to the desires of the large corporate clients of Jones Day -- the very clients that the TIP is positioning itself to serve.

Can one believe that the first set of these unaccountable bylaws were completely divorced from the interests of Jones, Day, Reavis and Pogue's present and future clients? For example given that Jones Day's intellectual property practice is among the largest in the world, it is likely that many of them would not like to have ICANN approve new TLDs since they will then have to defend their marks. And is not there at least the danger of a conflict of interest in Jones, Day being likely to represent both ICANN and those who may want to challenge it? It is troublesome to know that Sims, who is such a prominent player of late in this negotiation, comes from a law firm that would stand to benefit enormously if the ICANN deal goes through unchanged. And that Sims, having been just the pro bono attorney, with Postel gone, appears to be the principal with whom others are negotiating. Maybe part of the final agreement should be that the legal contract for ICANN representation should be put out for bid?