News for hackers. Stuff that matters.
Re:Facebook Tries to Make Violations of Terms of Use Into Cr... by jacquesm on Tuesday May 04, 2010 @09:02PM (#1319978)
This is all extremely interesting, especially the fact that you've been able to narrow down the 'range' of possible values by some real world analysis.

So, if I understand your piece correctly the value of the power changes over time depending on the size of the network and the available pool, in other words even the log(n) factor is not exact, in the beginning it is probably closer the to the original n^2, whereas later on it moves to more sedate territory, but it always seems to be higher than the '1' added by attaching another node to the network.

The 'competing with ebay' has a nice counterpart. In the netherlands there was a small local site called 'marktplaats' that had entrenched itself in the early days of the web, and nothing ebay did to dislodge it worked, so they ended up buying it.

The kicker is that 'hyves' (the dutch counterpart to facebook) is losing ground because people have a lot of international connections to friends and family, but trade seems to be limited to geographic boundaries and hence marktplaats succeeded where hyves is in trouble.

Re:Facebook Tries to Make Violations of Terms of Use Into Cr... by btilly on Tuesday May 04, 2010 @10:53PM (#1320191)
The Netherlands examples are interesting.

In the piece I quoted, it isn't that value scales like nn then like n log(n). The argument is that value scales like n log(n), which for small n looks more like nn than it does for large n. That said the argument given implicitly assumes that most of the people in the network have most of the people they really care about in the network as well. This assumption is less likely to be true for small networks than large ones.

It didn't make it into the paper, but an interesting optical illusion was discussed while we were writing it. Most of us judge the ubiquity of a network according to how many people we personally know who use it. Therefore people's value from belonging to a network tends to scale linearly with the size they perceive that network as having. (A typical American neither perceives nor values the growth of that network in China.) Which means that each individual observes what Bob Metcalfe did when he came up with his rule of thumb. We found that interesting, but left it out of the paper.

Another discussion we had was on different networks that scale differently. The two extremes we came up with are the telegraph networks and eBay. In the case of the telegraph network there is an asymmetry between senders and receivers. It is easy to make every resident of a major US city a potential receiver, at which point the network value scales linearly with the number of senders. And at the opposite end, auctions scale in a less approximately linear fashion. However we decided against including that because we didn't have detailed enough data to support our preliminary conclusions.

Also you should note that the IEE Spectrum article is an abridged form of http://www.dtc.umn.edu/~odlyzko/doc/metcalfe.pdf. The full version offers several different lines of reasoning that all converge on n log(n) scaling rules.