Santa Fe

Late Tuesday night, I arrived in Santa Fe. I checked in and walked down to the terminal room that had been set up for the IETF. After reading several dozen mail messages, and with over 100 still left to wade through, I didn’t have the heart to continue and called it quits.

Standing in the hall was Paul Mockapetris, the inventor of the Domain Name System. Paul and I left the hotel to go find some dinner, choosing a restaurant stochastically. The waitress came over, and gave us the introduction obligatory in such chi-chi places.

“Hi, I’m Foo and I’ll be your waitress this evening.”

She handed us each a flimsy little card. On the card were the words “fish,” “chicken,” “beef,” and “lamb.” A little icon was next to each word. There were a couple of other icons signifying potato and onion ring. No prices. No prose.

“What is this,” Paul asked, “a RISC menu?”

I decided to see what the wine list would look like in a place like this. Foo came back a few minutes later carrying four bottles of Chardonnay and plopped them on the table. We pointed to our desired icons, Paul pointed at a bottle of wine, and our point-and-click supper was on its way.

The next morning, suitably fortified with a breakfast of a burrito stuffed with Chipolte cactus and eggs, I followed a stream of what were obviously computer engineers to the morning plenary.

The IETF differs from most other industry meetings in that it is not a show or a conference. It is a place where people come to work. The plenary was very short and consisted of a technical presentation on ATM. Then, people went off to their working groups. Working groups are the main business of the IETF. It is in these meetings that people get together to develop the standards that make the Internet work.

The standards that have come out of the IETF have been impressive. They are produced quickly, and they work. In the area of network management, for example, the IETF produced an object identification hierarchy, management information bases for a wide variety of modules, and the Simple Network Management Protocol.

The IETF has been remarkably successful, but as it grows in size, its ability to accomplish work appears to diminish. Veterans of the IETF, like Marshall T. Rose, are vocal about the negative impact of success on the ability to accomplish real work. Marshall likes to distinguish between two types of attendees: the “goers” and the “doers.” The goers are those who like to go to conferences, forgetting that the IETF is a working group, not a conference.

The basic difference is one of self-definition. Somebody like Marshall T. Rose defines himself in terms of accomplishments: “the father of ISODE” or “a leader in the development of SNMP.” Others, however, define themselves by position: “head of a working group” or “member of the ACM.”

This conflict has a fundamental impact on groups like the IETF. The goers try to get themselves named as working group chairs, and all of a sudden committees for the sake of committees start to flourish. The number of working groups topped 60 by the time of the Atlanta meeting in Summer, 1991. Stev Knowles, the most vocal minority of the IETF, suggested a working group be formed with the purpose of reducing working groups. Somebody else suggested that a working group be formed to study the question.

Meanwhile, it keeps getting harder to get real work done. People attend meetings who have not done the preliminary reading. Working group sessions get stalled with naive questions. Even posting required reading lists to the net doesn’t seem to get people to do their homework (or to attend and keep quiet if they hadn’t done the reading).

Marshall T. Rose grew tired enough of the “goers” that, over a drink, he suggested a radical system of Certified Protocol Engineers. By Marshall’s scheme, in order to attend a working group meeting and participate, you must be certified by a board of your peers in that area. Network management, routing, and mail systems, are all examples of possible areas.

The board in an area would be bootstrapped by two people of unassailable quality, who would draft the oral and written exams. Once certified in the area, you would be part of the governing group. The system would be quite similar to that used in the medical profession.

Working group participation would thus be limited to certified protocol engineers. Anybody could attend a meeting, but to participate you must either be certified or be invited by the chair. This fairly radical proposal would probably never get passed — and was certainly proposed with tongue planted firmly in cheek — but it makes one think about how such a technocratic priesthood might function.

Aside from the usual work of readying router requirements and massaging management information bases (MIBs), the IETF has a ritual bloodletting plenary session on the penultimate day. The plenary always starts very slowly, with IETF chair Phill Gross uttering the obligatory platitudes of introduction.

While he is speaking however, you can look around the room and see people positioning themselves at the microphones, waiting for the meeting to open to the floor. Usually, if the controversy will be especially loud, you can hear rumors in the halls in the days before the plenary.

The Sante Fe meeting, however, had seemed fairly quiet and nobody was expecting a major controversy. Lack of real issues, however, will not always stop people from speaking.

As soon as Phill Gross stopped speaking, one engineer stepped to the mike and started raving about the Internet Activities Board. His proposal was to abolish the board as superfluous.

I was sitting in the back of the room, looking over the shoulder of Jon Postel, editor of the RFCs and a long-time member of the IAB. Jon has a habit of scrunching down in his seat as he gets more and more disgusted. As the engineer continued to rave, Jon was almost sitting on the floor.

Finally, people started to chime in with the monologue, interrupting the mad ravings of the lunatic engineer and reminding him that the IAB may have problems but it certainly had played a valuable role in the development of the Intemet. Jon Postel slowly started sitting up in his seat.

The IAB doesn’t always get accolades at the IETF meetings. The role of the IETF is technical advisor to the IAB, which goes ahead and makes policy decisions. In the previous IETF meeting, held in Atlanta in July, tempers had flared as high as the humidity.

The story of that firestorm helps illustrate some of the inherent conflicts in the process. The IETF had developed an MIB for SNMP-based management of Ethernet modules. The working group that developed the MIB included many of the standard IEEE-developed management variables, but also included a few optional variables that the group felt were necessary.

After several meetings and much e-mail correspondence, the working group forwarded the MIB to the general IETF. At a plenary meeting, the MIB was put on the table and no objections were voiced. It then went to the IETF Steering Group (IESG), composed of leaders from each of the main areas the IETF works in.

The IESG examined the MIB, saw no objections, and forwarded the document up to the IAB with a recommendation for approval. This is the normal process that any recommended standard originated by the IETF takes.

Once the document hit the IAB however, it sat. People were busy and had lots of things to do. Finally, it came time to act on the document. Tony Lauck, the chief network architect for Digital Equipment Corporation and a member of the IAB, looked at the MIB and thought that some existing Ethernet vendors would not be able to easily implement the standard. He felt that it was different from the IEEE standard and that this was not necessarily a good thing.

This is all well and good. The role of the IAB is technical arbiter of the TCP/IP protocol suite and Tony Lauck had identified what he saw as technical issues. He marked up the document, crossing out some variables and adding detailed instructions for revision of the MIB. The IAB then sent it back down to the working group.

The working group exploded. They felt that there had been many opportunities to provide technical input into the process and that the MIB reflected a technical consensus consistent with the network management framework that had become an Internet standard. They saw the IAB as making a crassly commercial decision, caving in to the wishes of a few vendors instead of providing leadership in the standards arena.

In Atlanta, everybody came to the microphone and started giving their views. Why hadn’t the IAB attended the working group meeting if they had concerns? Why had the IAB made technical edits instead of providing policy guidance?

Karl Auerbach of Sun Microsystems got up to speak. A lawyer by training and an accomplished engineer, he is a vocal participant in the IETF and an active SNMP developer. The rewritten MIB had been discussed at an unscheduled working group meeting the previous evening, and Karl was concerned that this violated the open, consensual nature of the IETF.

Others got up and complained that the IAB had issued several different explanations of what had happened, each different. Others were concerned that the IAB had tried not to tread on the IEEE turf, neglecting technical leadership for politics.

Finally, members of the IAB admitted they had fumbled the question. The issue was not technical, it was procedural. How to run an informal standards process like the IETF, yet preserve due process safeguards, has been a continuing problem as the body grew in size from the original 13 participants to the several hundred attending at Atlanta.

The issue of due process was certainly a crucial one. It appeared to some members of the IETF that Digital Equipment had sabotaged a standards action on commercial grounds. Karl Auerbach raised the question of the potential legal liability of the group.

The questions of accountability and liability had not escaped the attention of people like Vint Cerf, then chair of the IAB. He was actively working to promote a new professional society, the Internet Society, which would sponsor the activities of the IAB and appoint the members.

The IETF was fun, but by the time an angry mob got around to putting Jordan Becker, vice president of ANS, on the grill about the transition from the Tl to the T3 NSFNET backbone, I was ready to leave. Seven weeks on the road had certainly been enough for me.