Many networks consist of clusters of endpoints knitted together by many short links and a much less dense set of long links which tie clusters together. A disproportionate share of the value of a network to its participants flows through the long links. Here are four examples with practical implications.
Social Networks
The short links are between us and the friends and family we see and/or communicate with frequently. The long links are between people who don’t communicate very frequently. Whether it’s a new job, new spouse, or new business opportunity you’re looking for, it’s much more likely that you’ll find what you need through one of your relatively few long links (or the long link of a friend) than through one of your many short links.
Why? Because you and your cluster of short link friends all pretty much know the same other people (that’s why you’re a cluster), know about the same job opening and business opportunities. Your long link acquaintances know a lot of different people and different things than the people you talk to all the time. So, if you need something totally new, it’s likely to be found via a long link
The famous (unprovable) assertion that no two people on the planet are separated by more than six links is actually an observation on how important long links are to human interaction. If we were just all connected to a random selection of other people and clusters didn’t have a pronounced tendency to grow and use distinct long links, there would be many more hops to get from one endpoint (person) to another.
It also turns out that some people specialize in being the “long link gateway” for whatever clusters they’re a part of. Most of us have relatively few long links and rely on our better networked friends when we need a good, new connection.
The Transportation Network
I’m reading Wedding of the Waters by Peter L. Bernstein – a great book about the building of the Erie Canal. Bernstein summarizes a report US Treasury Secretary Albert Gallatin wrote in 1808 on the need for a long distance transportation network in the new United States:
“Most localities in the 1800s… had little interest in trading with their neighbors, who… were producing the same types of goods and services. Consequently, Gallatin explains, the first pathways of communication built in an undeveloped economy like the young United States were destined to be unproductive. In order to do any volume of business, connections over long distances were absolutely essential. The urgent need, Gallatin explained, was for a nationwide network of roadways and canals linking each local pathway to the whole.”
In other words, you gotta have the long links. Read the book to see how spectacularly the Erie Canal functioned as a ong link affecting not only the economy of New York State and the US but also of parts of Europe.
The Electric Power Grid
There’s an article in today’s NYTimes about huge costs imposed on electricity users by congestion on the electrical grid. Much of the competition which should have resulted from deregulation of the power industry isn’t realized because you can’t get power from here to there when you need it. Expensive local peaking facilities are used while more remote but cheaper facilities have extra capacity.
Long electric links need high towers. No one wants the towers in their view.
Yesterday I blogged about decentralized generation of electricity. Some of the time decentralization will reduce the load on the long links. However, long links make decentralization more effective as well. Huh? Well, suppose we’ve all got a lot of windmills here in Vermont but the wind isn’t blowing here today. Sure be nice to get some power pumped in from that big storm out in Chicago that’s got the turbines spinning. Every node on a network gains from being connected to a big grid with adequate long links!
The Internet
Until the existence of the Internet, long links were universally more expensive to use than short links. It wasn’t long ago that we paid for phone calls not only by the minute but also by the mile. Transportation costs (but not always prices) are distance based. We used the long links because they were valuable enough to justify their higher cost.
But distance doesn’t exist in Internet pricing. Email costs the same (usually no incremental cost) whether it’s going across the street or around the globe. In fact, an email may travel practically around the globe to get across the street and no one cares. If it weren’t for some obsolete tariffs and vestigial monopolies who own endpoints, phone calls would all be priced like email. Of course, voice which stays on the Internet isn’t priced by the mile or even the continent of destination.
We can access websites anywhere as part of the cost of our local connection.
So, pretty much for the first time in human economic history, the long links of communication don’t cost any more to use than the short ones even though they’re much more valuable. No wonder there’s so much opportunity.
This series on distributed networks begins with this post.