5. END-USER’S INTERNET DEMAND FOR ACCESS AND USAGE
5.4. S OME S TATISTICS O N I NTERNET A CCESS
Mark Lottor, who at one time worked at the Defense Data Network (DDN) Network Information Center (now InterNIC42), provides the basis for most of the statistics published about the Internet. Specifically, he has published statistics with the number of hosts on the Internet (world-wide) per year since 1981. Lottor shows that there were 6,642,000 hosts on the Internet as of July 1995. Rickard (1995), states that 64% (4,228,207) of those hosts were located within the US.
One important remark has to be made. Hosts are not necessarily end-users. In general, a host is a computer with an IP address, permanently connected to the Internet. On the one hand, when someone uses a dial-up connection, he obtains a temporary IP address and turns into a host for the duration of the connection. On the other hand, a university server (host) is permanently connected to the Internet and has a permanent IP address (but these hosts are continuously shared by several users). Lottor’s data was collected by computing the number of hosts that would be connected to the Internet at any particular point in time (both
temporary and permanent addresses). As is apparent, this computation requires some adjustment factor to account for multi-users: Those sharing hosts. This is particularly important when observing that the .edu domain (i.e. colleges and universities), for example, is the second largest users group on the Internet. These institutions typically have a ratio of around 10 students per Internet host. However, as it is shown below this adjustment factor seems to be much lower for the U.S. as a whole.
42 InterNIC--Internet Network Information Center.
5.4.1. Exponential Growth
Lottor’s data, when plotted, allows a display of the exponential growth that everyone talks about (see Figure 10) and which is mainly the result of: The industry—the wide adoption of
optics, government promotion, and technology innovations (such as Domain Name Service).
Figure 10. Exponential Growth Of Internet Hosts.
Aug-83 Oct-84 Oct-85 Nov-86 Dec-87 Oct-88 Oct-89 Oct-90 Oct-91 Oct-92 Oct-93 Oct-94 Jul-95 Jan-96
0 2,000,000 4,000,000 6,000,000 8,000,000 10,000,000
Aug-83 Oct-84 Oct-85 Nov-86 Dec-87 Oct-88 Oct-89 Oct-90 Oct-91 Oct-92 Oct-93 Oct-94 Jul-95 Jan-96
Source: Mark Lottor, Network Wizards 1996.
5.4.2. Number Of Users On The Internet
O’Reilly’s (1995) is one of the most accurate ones as far determining the number of users on the Internet. The results were published in September 1995. The survey’s methodology is quite accurate because:
• It is based on a random sample
• It is a large scale sample
• It is independent of other Internet derived information
This survey concludes that there are 5.8 million adults in the US accessing the Internet without using a commercial on-line service. Therefore, with the estimate of the number of
(or computers) on the Internet in the US (Lottor’s data), and by noticing that these two estimates were computed at similar periods in time, one can determine the ratio of users per host on the Internet in the US. The result gives 1.37 users per computer in the US. To expand this estimate worldwide, we would multiply this factor by the number of hosts on the Internet world-wide (6,642,000) to give 9,111,096 users on the Internet world-wide as of July 1995. There are some flaws to this calculation, but it turns out that is one of the most accurate ones. For example, it is expected that the number of users per computer outside there is much greater than inside the US. Also, the number of users connected through a commercial on-line service is not included here (Rickard, 1995).
An additional speculative prediction that one can make is about the number of users on the Internet as of January of 1996. By using the same users/computer ratio (1.37) and
multiplying it by the number of host from Lottor’s data (9,472,000), the conclusion is that there were 12,976,640 users on the Internet world-wide as of January 1996. However, because of the unknown and potentially unlimited numbers of multi-user computers and network or application gateways, as well as the existence of innumerable temporarily connected, non-advertised, or firewall43 protected machines, it is not possible to establish a complete total or correlate any of this information with the number of end users with great accuracy.
5.4.3. Access Locations
Another survey, the Nielsen’s study44 was published on December 1995. This study found that while home Internet connections are important, locations other than home were
significant sources of access. Reportedly, 62% of the users said that they had access at
43 See glossary for explanation.
44 See http://www.nielsenmedia.com.
home. Interestingly, the research showed that 54% of the users had an Internet connection at work and 30% had access at school (as the sample consisted of persons 16 years and older, at school is synonymous with college in most instances). On average, each Internet user had 1.4 different types of access locations.
Table 2. Access Locations
Nielsen FIND/SVP
Home 62% 69%
Work 54% 47%
School 30% 21%
Yet another study conducted by FIND/SVP,45 a large information services provider, reported that nearly half of all Internet users (46%) use a commercial on-line service to access the Internet. Figure 11 below shows that America On-line was the preferred Internet access method, although nearly a quarter of all users say they have more than one way to get on the Net:
45 See http://etrg.findsvp.com.
Figure 11. Internet Access Method46
America On-Line Work Server Academic Server ISPs Compu-Serve
Prodigy MicroSoft
30%
28%
27%
25%
11%
9%
2%
America On-Line Work Server Academic Server ISPs Compu-Serve
Prodigy MicroSoft
MicroSoft Prodigy Compu-Serve ISPs
Academic Server Work Server America On-Line
Source: FIND/SVP, 1995.
5.4.4. Computer Type
GVU447 reports that the most common Internet access platform is Windows (62%), followed by Macintosh (21%) and Unix (7%). By comparison, the NPD Group, Inc.48 reports that Windows-based PCs now account for 59% of household PCs, while a Macintosh is found in 14% of all U.S. homes.