One frequently finds newspaper or magazine articles about the Internet or the World Wide Web stating that the number of servers on the WWW is doubling every 53 days, "according to a source at Sun Microsystems." Well, I am that source, and I don't believe in the 53-day estimate any more.
Unfortunately, it is proving to be almost impossible to stop writers (whether journalists or Usenet posters) from citing this number: it has taken on a life of its own, as people cite it based on secondary sources (earlier articles) instead of primary sources (direct interviews with me). The 53-day estimate for web-doubling has thus proven to be a true "meme": a mental concept that multiplies by spreading through the population and grasping hold of the imagination. Encapsulating the Web phenomenon in a single number is a powerful idea and is much easier to communicate than more elaborate explanations based, for example, on modified Bass curves for innovation dissemination and technology transfer. As a human factors expert, I guess I am usually satisfied when I can make complex things simple, but we should remember that sometimes, things are more complex.
When I was writing my recent book on hypermedia and hypertext, I needed a simple way to describe the growth of information for the chapter on information overload. One of my estimates was that the number of available servers on the WWW was doubling every 53 days. In addition to putting this estimate in my book (where it is supplemented by rather elaborate diagrams and other ways of analyzing the situation), I also used this estimate in a number of talks I was giving about online publishing, and several other Sun speakers did so too. In reporting on these talks, many writers quoted us as saying that Sun estimated a 53-day doubling of the number of WWW servers, and the meme was born.
I arrived at the estimate by quite simple means: I collected estimates of the number of web servers from many public sources, such as the various WWW robots (programs that traverse the Web and move on from each server to any other servers it links to). I used the average of the numbers from those robots I found credible (admittedly a subjective judgment) as an estimate of the number of servers on the Web in any given month. By the beginning of 1995, I had been collecting this data for almost one and a half years and I was thus in a good position to estimate the growth rates represented by the monthly estimates of number of web servers. In January 1995, the growth rate corresponded to an annualized rate of about 12,000 percent, meaning that the number of web servers would double every 53 days.
Of course, growth rates this fast cannot continue indefinitely, and now (March 1997), the same method provides an estimate of 333 percent annualized growth, corresponding to a doubling of the number of web servers every 173 days. The number of web servers is growing ever more rapidly in absolute terms, but the annualized growth rate will probably keep going down until at some point of time it approximates the growth rate of the Internet as a whole, which has been stable at about 100 percent per year ever since the Internet was founded.
Growth curves for the Internet and the WWW on a logarithmic scale.