Weighing the Internet

the-dark-kangaroo writes "Jason Striegel has taken Physics to a new dimension by 'Weighing the Internet.' Well, actually calculating the total number of users online in one day. The conclusion that was reached was that there are ~519 million users per day online. Also, 'From what we calculated, it would appear that roughly 41 percent of internet users did not log in that day.'"

Woah, not even close

[kyndig]

This is so horribly full of conjectures, uncontrolled data resources, and just pure speculation. The figures are based off Alexa Toolbar users, and one website visitor ratio. The author uses these as the base of forumlating a simple division/multiplication approach to postulating the gross users of the internet.

Suggestion for more accurate collection of information. Talk to ICANN or that nifty website senderbase.org [senderbase.org] that has a broader view on traffic flow across the internet.

2600 Hz

[Anonymous Coward]

That was the "operator" login for telephones.

Interesting Conclusions

[chrisp9446]

Wait a second...didn't we conclude yesterday that 1/3 of all studies are bunk? Well, at least these guys did admit their data wasn't statistically valid ;).

Re:Still Logging In? The System Isn't Finished.

[photon317]

Linux is already capable of booting extremely fast, but it's the distro guys that are lagging on making it happen. Basically, a large part of the boot time is starting a bunch of services sequentially. However, if you have proper service dependency information (like LSB-based distros should all have, and Gentoo has for sure), instead of just boot order numbers (/etc/rc2.d/SNNsomeservice), you can parallelize a lot of the boot process. Add to that the fact that except for kernel upgrades, you don't really need to reboot linux anyways, and 2.6 has integrated software suspend to HDD, and you can boot even faster by just suspending to disk instead of shutting down.

Re:horribly ?

[image77]

Attraction vs. waist size would be more representative

Re:Continues on...

[Anonymous Coward]

Actually, I did continue the study. What I went on to say was this [blogcadre.com].

- Jason

It can be measured, but not like that

[louarnkoz]

The methodology presented here is deeply flawed: it extrapolates a large number based on a very small sample and on unsupported assumptions about browsing habits. Yet, it is possible to actually measure the number of users with some proper method.

The most obvious method is a basic opinion poll. Take a large enough random sample of the earth population, ask simple questions like "have you used the Internet ever, this year, this month, this week, today", compute the average and extrapolate.

In practice, taking a world-wide poll is not very practical, but it is certainly possible to perform polls on a country by country basis, and then compute the results. In fact, such polls are regularly conducted, and the results are just a google search away, at least for major countries.

Polls are snapshot at a moment in time, and this is problematic. If you don't pay attention, you end up adding the number of users measured in China last January, in the US last month, in Finland in May, etc. So, you want to complement the polls by an indication of trend, something that you can easily measure at frequent interval.

One possibility is to use Internet host counts, which can be obtained by sampling the DNS (see the Internet Domain Survey [isc.org]). One can measure the number of host in a country and the number of users at the time of the poll, the current number of host in the same country, and extrapolate.

There are other potential sources, e.g. measure the volume of traffic, the number of dial-up and broadband subscriptions, etc. Again, it is possible to link these numbers to various poll data, and maintain estimates.

By the way, the Internet Domain Survey in January 2005 showed 317.6 million IP addresses in use. The typical broadband connection uses one IP address per household, i.e. for 1 to maybe 4 or 5 users. A dial-up connection typically only use an address only a fraction of the time, so the ratio is even higher. Then, there are about 650 million PC available worlwide, many of which are shared. Based on that, there were probably somewhere between 500 millions and a billion users on the Internet.