IETF’s quest to be quicker moves slowly

MINNEAPOLIS – Eighteen months after acknowledging pervasive organizational and productivity problems, the Internet Engineering Task Force continues to search for remedies designed to bring important network standards to market sooner.

A pivotal debate at last week’s IETF meeting in Minneapolis failed to provide any magic bullet, but it did produce a new sense of urgency and resolve to get the problems fixed in the year ahead.

In the past 18 months, the IETF has created several committees to look at reorganization, standards tracks and administrative structures. These groups are publishing final documents that outline the IETF’s consensus on what its problems are in each of these areas.Having come to agreement on what the group’s problems are, the IETF will now try to gather the consensus of its 2,000-plus participants around the steps required to fix these problems. The debates from last week will continue on the group’s e-mail mailing list for several months. The IETF will host a follow-up discussion on possible solutions at the group’s next meeting in February in Seoul, Korea.

Founded in 1986, the IETF has created many of the communications protocols that power the Internet – from behind-the-scenes routing techniques to the e-mail infrastructure to the ubiquitous DNS. The IETF is known for its brilliant but idiosyncratic participants, who clash publicly and vehemently over the engineering trade-offs required to create industry standards.

In recent years, the IETF’s all-volunteer leadership has been burdened with the volume of protocol proposals it receives as well as with the informal structure of the organization that was created when the Internet was more of a research project than a production network. Several of the group’s area directors have resigned citing overload.

Meanwhile, standards development work has slowed to the point that it often takes more than five years to finalize specifications that network hardware and software vendors are looking to deploy.

“We need to move away from being stuck and get back to producing quality output for the Internet community,” says IETF Chair Harald Alvestrand, an engineer at Cisco. “Producing bad standards quickly is not the goal.”

A tougher challenge for the IETF – and one that’s largely outside the restructuring debate – is how to change the group’s divisive culture and reduce the infighting among its participants, who often don’t read the documents they argue against. Many working groups have gotten so mired in philosophical debates that they miss their deadlines, produce poor-quality documents or fail to produce any documents at all.

“For a standards body, an engineering body, to have this level of debate about issues from people that have no knowledge of these issues is the most serious problem,” says John Klensin, a former Internet Architecture Board (IAB) chair whose involvement with Internet engineering dates to the original ARPANET. “If people are commenting with great vehemence on documents they haven’t read, that’s going to kill the group. Nothing else really matters.”

Alvestrand says the IETF’s main problems are that it lacks a common mission, it isn’t developing standards in a timely fashion, and its management structure hasn’t scaled to accommodate its workload.

“The IETF’s structure is like an informal gentleman’s agreement,” Alvestrand says. “That structure isn’t working very well. We’re wasting too much time deciding who does what.”

A new approach

Alvestrand proposes that working group chairs – about 250 people – take on more responsibility to steer development efforts and keep them on schedule. He also advocates a new document review process to ensure that experts on security and other matters examine draft standards before they are presented to the IETF leadership for approval.

“We have to move caseloads away from [the IETF leadership],” Alvestrand says, pointing out that the group received 900 documents in the four weeks leading up to this meeting’s cut-off date. “We can’t expect to organize 900 document reviews in that time frame. Serious documents need to have several people review them.”

The IETF reforms likely will expand the role of its sister organization, the IAB. The IETF and the IAB are chartered by and receive funding from the Internet Society, which has a goal to assure the evolution of global Internet and internetworking technologies.

“The IAB will likely be given more responsibility for document review,” Alvestrand says. “They are the source of significant technical expertise.”

Patrik Faltstrom, a former IETF Applications Area director and a current member of the IAB, says the group’s rank-and-file participants seem to support the restructuring efforts.

“We’ve been talking about this for 18 months, so people are pretty comfortable with it,” Faltstrom says. “People understand that change is needed.”

Money matters

The IETF also suffers from attendance and financial problems. About 1,200 engineers gathered at last week’s meeting, one of the IETF’s three yearly gatherings. That’s down from 1,400 at the group’s summer meeting in Vienna, Austria. At the peak of the Internet boom three years ago, the IETF attracted 2,900 people to standing-room-only sessions.

Because attendance is off, the IETF is losing money on facilities it has to book months in advance.

“Meeting fees are lower than expected,” Alvestrand says. “We have surpluses from previous years [to cover the shortfall], but this is a significant problem that has to be addressed.”

One option for the IETF to address its financial problems is to incorporate and take charge of its own budget. The IETF receives about a third of its funding from the Internet Society, and the rest comes from meeting fees.

In 2002, the IETF slipped into the red when its meeting attendance began dropping. Alvestrand projects that the group’s surplus funds will be used up in 2004 unless it can cut costs.

IETF leaders are having difficulty trimming their expenses because they don’t control the logistics of their own meetings and can’t put them out to competitive bid. Instead, those meetings are coordinated by Foretec Seminars, a for-profit arm of the Corporation for National Research Initiatives. (CNRI) Similarly, CNRI provides the IETF’s administrative support services. Foretec and CNRI provide these services under informal agreements that date to the IETF’s founding.

Many long-time IETF participants say the group needs to speed up its processes to retain the enthusiasm of top engineers, produce higher-quality specifications and encourage companies to commit their staff to the group’s efforts.

Transport Area director Allison Mankin, a Bell Labs researcher who has been involved with the IETF since 1987, favors giving more authority to working group chairs and cross-training them on different aspects of the Internet.

“We have huge numbers of really talented people, and many of them are in niches where they don’t get the broad experience that we need,” Mankin says. “One way they can gain a broader perspective is to exchange critical information with others. Working group chairs in Security and Transport are not likely to go through documents together.”

Mankin and other IETF leaders say they are losing the ideas and energy of good network engineers who are fed up with the group’s slow processes.

“We need to speed up our processes because they make people frustrated and they start working on other tasks,” says Faltstrom, a Cisco engineer and 10-year IETF participant.

“No one has the energy to follow these working groups until they end. Then the final review of the documents is not as good quality. We will get better quality having focused work that’s done on time,” he adds.

Mankin says it’s more important for the IETF to produce quality specifications than it is to produce speedy specifications, despite market pressures.

“Time and again, people say the IETF is going to be irrelevant if it doesn’t move quickly on something. But then the quality specification comes out and the market adopts it,” Mankin says.

Alvestrand says he hopes to finish the IETF reform process before his term expires in March 2005.

“I’m hoping to leave the group in a place that’s much better for my replacement,” he says.