Articles Weblogs Books School Short Cuts Podcasts  

Peer and Web Services are Technologies of Connection and Coordination

by Jon Udell

Imagine a technology that enables you to reach across the network into my file system, or that reflects a change in my document directly into a cell of your spreadsheet. I watched these kinds of demonstrations at the O'Reilly Peer-to-Peer and Web Services Conference with a curious sense of deja vu. I had seen all this before, almost a decade ago, in Windows for Workgroups. So what's really new here, beyond a protocol facelift that replaces SMB with WebDAV, and NetDDE with SOAP?

The scale and richness of cyberspace changes everything. Now that the Internet encompasses all of those LAN-based workgroups and so much else too, new modes of collaboration can take root and flourish.

There was a remarkable energy at this conference, and it drew an amazingly diverse bunch of hackers, lawyers, scientists, soldiers, journalists, and policy wonks. Like the proverbial blind men, none of us can yet clearly discern the shape of the elephant we are collectively exploring. But we are starting to catch glimpses of the thing, and here are some of mine.

Group dynamics

Metcalfe's law says that the value of the network grows as the square of the number of nodes. But as Clay Shirky pointed out in his keynote address, the value of a group-forming network rises exponentially with the number of groups formed. Finding ways to harness the power of that combinatorial explosion is the major challenge that lies before us. Because the notion of a group is so fundamental, it's wildly diverse. All sorts of things can form groups -- including CPUs, software processes, storage devices, digital assets, and people. Groups range in size from two on up. They may persist for seconds, days, weeks, or years. Their memberships may remain stable over those periods of time, or fluctuate.

Related Reading

2001 P2P Networking Overview2001 P2P Networking Overview
Clay Shirky, Kelly Truelove, Rael Dornfest & Lucas Gonze
Sample Chapter
Full Description

In a session on military uses of P2P, Earl Wardell -- who is helping the joint chiefs rethink hierarchical command and control -- described a fleeting encounter between two US planes over Kosovo. One, a fighter jet, was spotting for the other, a bomber. Because the two pilots couldn't form an ad-hoc and secure network of two nodes, they were forced to talk on an open radio channel, were overheard, and spooked their target. In another session, Adam Gross -- representing Grand Central Networks, whose Web services network joins companies trying to interconnect business processes with partners -- said that "integration wants to be free." Of course, there's no free lunch; it always costs something to form communicating groups. What's becoming clear, in this next phase of the Internet's evolution, is that we must identify those costs, and begin to think architecturally about how to reduce or reallocate them.

The theme of biology, introduced in the keynote, resonated throughout the conference. We see interacting groups at every level: organelles, cells, organisms, societies. Each is bounded by what Shirky called a "state horizon." It seems useful to ask how nature optimizes the coordination costs of systems that cross these boundaries; as one biologist pointed out, we don't yet have good answers to that question. Among the outcomes of this conference, I hope, will be some fruitful new collaboration between the information and life sciences.

Context and coordination

Related Material:

Reports from O'Reilly's P2P & Web Services Conference

O'Reilly Research's 2001 P2P Networking Overview Now Available

When group formation brings people together, many of the costs are familiar and easy to quantify. Can we find one another in the first place? If so, can we connect using compatible devices and protocols? Can we authenticate one another? Is there enough bandwidth? Can we communicate securely? These are well-known technical issues, and while progress may seem maddeningly slow at times, at least we know something about how to reduce these costs.

There is another kind of cost, though, about which we know much less. I call it the cost of context assembly, and it is the bane of all groupware systems. Think about customer service. If I'm having trouble using a Web service, my phone call or email message to a help desk agent can't say much about me or about the state of my troublesome application. In one of the most compelling software demonstrations I saw at the show, Internet Access Methods' Gerry Seidman inverted the help desk scenario. In a JXTA-enabled financial application, he clicked a Help button. "Now," says Seidman, "I wait for my phone to ring." Because the help request bundles application and user context, the screen of the service agent who would make that call can display the relevant pieces of application GUI and backoffice information.

The assembly of rich context, and delivery of it to an ad-hoc group at the instant it forms, is a capability sought in different ways, and for different reasons, by military researchers, enterprise application integrators, music-sharing enthusiasts, corporate knowledge managers, and just about everyone else, too. Solutions aren't evident yet, but there's a growing consensus about requirements, and that's good news.

Network topologies

The peer-to-peer application most in need of context enrichment is email, the original and still most successful groupware application. The point that email is a P2P application, both socially and architecturally, was made elegantly by Nelson Minar, former CTO of Popular Power. He evaluated client/server, ring, hierarchical, and mesh topologies along key measures including extensibility, information coherence, and fault tolerance. And he showed how emerging P2P systems that compose topologies can exhibit hybrid rigor. By combining a local client/server style within a global P2P mesh, for example, Morpheus does better on some key measures than Gnutella. Should we be surprised to find, therefore, that the topology of Morpheus is exactly that of Internet email?

Minar's point is that decentralization in and of itself means nothing. It's one style, centralization another. To be 100% pure P2P is as pointless as to be 100% pure Java or Perl or Windows or Linux. P2P and Web services are, some say, building blocks of an Internet operating system. Insofar as they define interfaces, broker connections, and allocate scarce resources, that's true, but it sounds oddly sterile. Perhaps the new technologies for group formation, content superdistribution, and collaborative filtering should be seen as adaptive responses arising from human colonization of cyberspace.

Jon Udell is an author, information architect, software developer, and new media innovator.

Return to