Computers with a view

Countless federal employees have unknowingly dabbled on their home computers with the technology that could change the way their agencies approach network computing.

Called peer-to-peer networking, it's the key to the popular instant messaging services offered by AOL Time Warner Inc., Microsoft Corp., Yahoo Inc. and others. It's also at the heart of Napster, the notorious online song-swapping service that has swept the music scene.

Both programs work by giving people a window into the computers of other people on their network. Instant messaging, for example, alerts users when friends or family are online so they can trade messages in real time. Napster, meanwhile, maintains an index of music files people have stored on their machines so that someone looking for a particular song knows where to find it.

These services represent a dramatic shift from traditional client/server computing, in which people access applications and data on a common server, yet work in virtual isolation. With P2P, most data resides on the net-work with a central system serving primarily as a broker, facilitating exchanges between people.

No longer isolated, people can tap into all of the resources available from their "peers" on the network. In short, the knowledge of one person on the network is equal to the sum of the knowledge of everyone else on the network. It's a collective strength unimaginable in traditional computing.

Even as commercial businesses continue to benefit from P2P, federal agencies now are exploring how they, too, can take advantage of the technology. But security concerns, regulatory roadblocks, bandwidth limitations and a general distrust of something new and untested have kept P2P from gaining a solid foothold in the public sector.

Brand Niemann, who did pioneering work on FedStats.gov and FedStats.net (see profile, Page 29), is a peer-to-peer prophet of sorts often called upon to explain the technology and its usefulness. FedStats is using Extensible Markup Language (XML) and P2P technology to enable more than 70 federal agencies using 200 statistical programs to draw on, add to and link to one another's statistical data "faster, better and cheaper," he said.

The FedStats.gov Web site, maintained by the Federal Interagency Council on Statistical Policy, provides a gateway to a wealth of statistics on economic and population trends, health care costs, aviation safety, foreign trade, energy use, farm production and more — much of which had previously been stored in paper form or in obscure electronic files.

Niemann, a computer scientist in the Environmental Protection Agency's Office of Environmental Information and a member of the FedStats Interagency Task Force, uses a recent Gartner Inc. report on P2P content networking that identifies five models of P2P (see box at left) to provide more details about this burgeoning technology. FedStats chose the data-centered model for its work on what might be the grandest example of XML-based P2P in the federal space.

FedStats is using Next-Page Inc.'s NXT 3 software, which includes a Content Network Protocol and Extensible Indexing Language, to power its data-centered system, where data is held on user's devices — desktop or portable — and dynamically indexed and made available to others when users connect with one another. Distributed files and content sharing are key applications, Niemann said.

The content "is much more than just catalogued and indexed Web links," he said. The data has "added value like hierarchical structure for improved browsing and XML tagging for improved searching." In addition, the content networks "don't just come and go at the whim of organizations, like some Web sites," because there is an emphasis on fostering and preserving collaboration as well as physical connections.

In the FedStats.net implementation — a test-bed Web site for exploring new approaches to disseminating statistical data and information — one NXT 3 peer server is used as the portal to several other NXT 3 peer servers, each of which allows file sharing from many FedStats agency Web servers. Those files are indexed and integrated into a system of folders that are labeled by agency and content subject or category.

"We connect devices together in a P2P framework and treat devices with huge amounts of content on them as a single [entity]," said Bruce Law, vice president of corporate marketing for NextPage.

Reports can be written and maintained by an agency on its own systems but still added to the FedStats content network. The system doesn't yet extend to portable devices, but CD-ROM content not previously available on the Web is included.

As the project grows and "we tag enough content on enough servers, the queries will begin to create tables and other data sets based on the request, eventually leading to a personalization feature that allows users to save the search and then redo it at a later time to get the new content that has been added since the last one," Niemann said.

In addition to the NextPage tools, FedStats is also using StatServer from Mathsoft Inc. and FileMaker from FileMaker Inc. for different components of the new site. Niemann is especially fond of the FileMaker relational database software, which allows users to share information with one another without going through a Web server as long as they supply an IP address.

Niemann demonstrated the FileMaker tool at an EPA users conference in Durham, N.C., in early April by having one of his assistants open a database on Niemann's computer at EPA headquarters in Washington, D.C., and mark it for sharing with any EPA IP address. Instantaneously, anyone at the conference or within the agency could access the database "without using a conventional Web server, just my Windows 98 desktop PC with a dedicated IP address," he said.

"The goal is to have as many agencies with their own content network nodes as possible," Niemann added. "As far as we know, this is the first use of XML P2P in the federal space, and if there are others doing it, it's not on this scale."

FedStats only uses publicly available data sources, so security is not a great concern. "There is no confidential or nonpublished data in the FedStats network," so making the information readily accessible for collaboration and editing only benefits the agencies involved, he said.

Law said that any federal programs that require information or research to be shared among several agencies could benefit from P2P technologies.

"Distributed research programs doing collaborative work on a topic from sources around the world can be created virtually and collaborated around the information," he said. "If five different agencies are working on a project, it could present a unified view of it without all the data [having to reside] in one place."

Law also said P2P could improve federal collaboration with state and local agencies because the content could reside locally but the partners could see a holistic view of the data, whether it's environmental information or documents shared by law enforcement agencies, and would not require manual updates from year to year. Niemann is working on the National Environmental Information Exchange Network to enable collaboration between the EPA and states, and to supply the agency with 90 percent of its data.

DARPA "Soldier's Radio'

The Defense Advanced Research Projects Agency is also experimenting with the potential of P2P. DARPA is using P2P technology for both networking and data-dissemination systems, said Paul Kolodzy, program manager of Small Unit Operations: Situational Awareness System in DARPA's Advanced Technology Office.

The networking project for soldiers on the battlefield has been designed to establish secure, reliable wireless communication while keeping overhead low for real-time communication using a "soldier's radio." The system uses mobile P2P nodes as intermediaries across long distances to coordinate both voice and data communications.

"It's akin to the Internet, where there are lots of servers between e-mails," Kolodzy said. "But different than the Internet is that our P2Ps are moving constantly to see what path is available to us — from one user to another — through the P2P intermediary system."

The soldier's radio is being developed by ITT Industries' Aerospace/Communications Division, which is working with MontaVista Software Inc. The network is based on the Linux operating system.

"With wireless, overhead is one of the driving mechanisms," Kolodzy said. "If half of your bandwidth, or communication capacity, keeps routing tables updated, you're wasting half the bandwidth and communicating only half as much data."

Traditional radio communications use as much power as necessary to reach the most distant user, but the P2P system offers power savings by using shorter links, which increases battery life. P2P also reduces interference and is "less vulnerable to eavesdropping from adversaries," Kolodzy said. "If you radiate less power...the less detectable you are."

DARPA's data-dissemination project aims to improve warfighters' situational awareness by giving them as much tactical information as possible about their area of the battlefield, such as the location of friendly forces and the movements of enemy forces. The system, which also has a text capability, works like Napster, which takes a request for a song and then identifies the best place for the user to reach it.

"On the battlefield...it's not a song," Kolodzy said. "It's a particular set of information geographically or organizationally oriented."

A query, which could include anything from logistical data to enemy or allied movements, is culled from a series of distributed servers addressed by the radio. For example, a soldier might use a voice command to announce that an enemy tank had been sighted at a certain coordinate. That information would be transmitted to soldiers in the area and automatically sent to other personnel when they move into that area, Kolodzy said.

The actual distances covered by these P2P systems are still being ironed out and depend on the data rate that the unit wants to use. High data rates mean shorter distances, and lower data rates mean more reliable communication across longer distances, with a goal of reaching beyond 50 kilometers, Kolodzy said.

The Army is the primary customer for both the mobile and mounted applications, but the Marine Corps and U.S. Special Operations Forces are also interested. By next spring, lab testing will be complete, and by September 2002, field testing and demonstrations of this technology will be finished, he said.

DARPA has been working on its P2P projects since before Napster and Instant Messenger burst onto the scene, but "the recent explosion has kept motivation high and been inspiring to us," Kolodzy said.

Feds Interested, But Concerns Abound

The FedStats program and the P2P projects being developed by the armed forces show that the technology is gaining momentum in the public sector. Nevertheless, Tim O'Reilly, founder and chief executive officer of O'Reilly & Associates Inc., a high-tech information provider, said he was surprised by the number of federal attendees at the company's P2P conference in February in San Francisco. In fact, because representatives from several agencies, including the U.S. Joint Forces Command, attended the first conference, there will be a "government track" at O'Reilly's next P2P conference in September in Washington, D.C.

Both NextPage and O'Reilly, as well as industry heavyweights Hewlett-Packard Co. and Intel Corp., are members of the Peer-to-Peer Working Group (www.peer-to-peerwg.org), a consortium that promotes the advancement of infrastructure standards for peer-to-peer computing.

"In general, P2P is so fundamental as a technology," O'Reilly said. "It is a way of people using their own resources" to improve communication and information sharing.

O'Reilly said some of the more interesting applications for P2P in the federal space include "distributed computational research in the supercomputing community" and document sharing of large databases for the legal, law enforcement and military communities.

But although O'Reilly said government agencies are interested in P2P, the recent wrangling over the legality of Napster highlights the unresolved regulatory issues that prevent some from embracing the technology. Another issue is accountability. Many in government are already wary of e-mail because it is auditable and messages can be requested and held for legal proceedings (President Bush sent his last e-mail in March for that reason), but there are no standards for P2P communications.

"There are also a lot of security issues that need to be hashed out, like how to keep logs and what should or shouldn't be included in them...and about communication between the corporate network with outside [machines]," said Quazi Zaman, advanced technology manager at Microsoft Federal.

One sign that some in the federal government are nervous about P2P came April 20 when the FedStats.net node was taken off-line. "As with all FedStats/digital government research collaborations, the exploration of peer-to-peer technology in the FedStats.net environment remains a proof-of-concept demonstration at this juncture," according to a message on the site.

Niemann was caught off guard by the shutdown, but he's still not betting against P2P. "Peer-to-peer technology has a role in integrating information and applications across the government...faster, better and cheaper."