Who's really winning the browser and Web server wars?
While Microsoft and Netscape slug it out on the client side, the Web server market is dominated by another player
An update to last year's column on browser market share with Chuck Musciano's predictions on who will dominate in the coming months. This time, Chuck also provides Web server statistics as well as scripts for monitoring users on your own site. (1,700 words)
Last year, around this time, we took a look at browser market share and its implications for HTML authors. It's time to revisit the browser market and address the server side as well, pinning down both the most popular browsers and servers in use today. The results are somewhat surprising, with lessons for both HTML authors and Webmasters.
This time last year, Netscape 2.0 was rising in popularity, displacing Netscape 1.0 as the browser of choice. Together, both versions of Netscape accounted for 90 percent of the browser market. Internet Explorer had eked out about five percent of the market, with the remaining five percent occupied by several dozen other browsers.
In last year's column, I showed how any Webmaster can determine which browser is being used by each visitor to his site. When a client connects to a Web site, it sends an identifying string to the server, usually indicating the name and version of the browser. The server can use this string any way it sees fit. At a minimum, most servers write the identifier to a log file; more sophisticated sites are reprogrammed to configure their HTML pages on the fly based on the browser type.
While servers vary, most servers based on the original NCSA
server write the browser identifiers to a file named
agent_log in the same directory as the
access_log. Since individual entries in this file can
be somewhat cryptic, I've made available a
script (see Resources below) that Unix users can use to convert the raw log entries
into more usable names. Once you've run the raw log through
sed, you can use the
uniq utilities to sort and count the various browser
types. When all is said and done, you'll have a nice neat list of
browsers, each accompanied by the number of times that browser was
used to access your site.
Current browser usage
For this year's analysis, I made a few changes to the script, adding references for Netscape 3.0 and 4.0, along with Internet Explorer 3.0. I then applied the script to my monthly agent logs and converted the resulting data into a fabulous chart. The data is based on an average of about 100,000 hits per month:
At first glance, the most noticeable trend is the recent growth in the use of Internet Explorer 3.0, seemingly at the expense of Netscape. To clarify this, I combined all versions of Netscape into one group and all versions of Internet Explorer into another, and created a second chart:
Indeed, Internet Explorer is showing a consistent upward trend, cutting into the market share owned by Netscape. From July to December 1996 my monthly data had shown Netscape owning about 85 percent of the browser market, Internet Explorer with 10 percent, and the remainder split among many other browsers. This finally began to change in December 1996, with Internet Explorer showing steady growth since then.
Keep in mind that Internet Explorer is probably not displacing existing Netscape users who are abandoning Netscape in favor of Internet Explorer. With the growth of Web usage, it is more likely that Internet Explorer is capturing more new users who have never before used a browser. This could be attributed to Microsoft's savvy bundling deals that placed Internet Explorer in the hands of more new users over the past few months. As the Web community grows, both Netscape and Internet Explorer are adding new users, with Internet Explorer adding them faster than Netscape.
What does this shift in browser share mean for HTML authors? The short answer is "nothing." While a growth in share for Internet Explorer from 10 to 19 percent may make for great Microsoft marketing, the reality is that Netscape retains close to 80 percent of the browser market. Unless your site demographics show a significantly different pattern, you should continue to target Netscape as the browser of choice when designing your pages.
Even with its current growth, Internet Explorer still does not
carry enough clout to influence the use of HTML on the Web. While
Microsoft may add unique capabilities to Internet Explorer, like the
<marquee> tag and rudimentary style sheet
support, none of it will catch on unless Netscape adopts similar
technology. Regardless of the marketing hype, Netscape is still the
tail that wags the dog in defining commonly-used HTML features.
Netscape's position as a market leader is only going to further solidify as Netscape 4.0 is released. I'll go out on a limb and predict that Netscape will return to a near ninety percent browser share when a stable version of Netscape 4.0 is generally available. Even with the current unstable pre-release, version 4.0 is already commanding three percent of the browser market.
The final release of Netscape 4.0 will also bring about major changes in HTML page design. Netscape 4.0's new layering capabilities will revolutionize page layout much the same way that the table features in Netscape 2.0 did. Further, Netscape's solid support of Cascading Style Sheets will finally bring style usage into the mainstream, something that Microsoft's buggy partial implementation never accomplished. The net result is that Netscape should continue to dominate the browser market and drive HTML usage for the remainder of this year.
The server side
While we're analyzing data and creating nifty charts, I thought it would be interesting to see how server usage is changing on the Web. Once again, Netscape and Microsoft are locked in a market share battle, trying to capture the majority of Web servers currently in use.
No one person can capture enough data to determine trends in the server business, since your server is not creating a convenient log of other servers' information. Fortunately, someone else is doing all the hard work for you, regularly polling several hundred thousand sites on the Web and asking the servers to stand up and be counted.
That "someone else" is Netcraft, a firm specializing in network security, Unix administration, and Web publishing. As a service to the Web community, Netcraft provides a site that catalogs the distribution of Web servers on a monthly basis. Each month, Netcraft automatically polls Web sites around the world and catalogs the servers being used on those sites. That data is rolled up into several different reports made available on the Web.
The number of sites polled grows each month. In April 1997, Netcraft polled 1,002,612 sites. It continues to seek out new sites and add them to its list. You can ensure that your site is on the list by asking Netcraft to examine your site. (See Resources below.) After returning the results, your site will be added to the list for the next month's survey.
The monthly reports provided by Netcraft are extremely detailed and slice the data in several different ways. I wouldn't presume to supplant its in-depth analysis, but found it interesting to extract a few top-level statistics. Feel free to visit Netcraft's site and get all the gory details; in the meantime, here is another delightful chart:
This chart shows aggregate usage for the five most common servers: Apache, NCSA, CERN, the Netscape family of servers, and the Microsoft family of servers. The Netscape group includes the Netscape Enterprise, FastTrack, Commerce, and Communications servers, along with the Netsite Commerce and Communications servers. The Microsoft group includes the Microsoft Internet Information Server, Microsoft IIS, Microsoft IIS W, Microsoft PWS 95, and Microsoft PWS.
I was surprised and pleased to see that Apache is, far and away, the most popular server on the Web. While the media is occupied with the marketing battle between Netscape and Microsoft, most Webmasters are quietly downloading Apache to run their Web sites.
It's easy to see why. Apache is loaded with features, is constantly updated by developers around the world, offers complete source code, and is completely free. Apache is a testimony to the spirit of cooperation and community that exists on the Internet. Without any compensation, hundreds of people constantly debug and enhance the Apache server and offer the results back to the Web. What more could you ask for?
Apache has held over forty percent of the server market for some time, and that trend shows little sign of changing. Recently, Microsoft surpassed Netscape's installed base, acquiring over 15 percent of the market. Netscape's current share is 12 percent. The only real losers in the current survey are NCSA and CERN, which are clearing dropping in popularity. That drop does not seem to be benefitting any commercial server vendor; instead, it appears that NCSA and CERN users are simply switching to the more robust and feature-rich Apache.
The popularity of a good, free server is easy to understand. While large sites have the resources to purchase a server and the desire for the dedicated support that a vendor provides, many Web sites are run on a shoestring. For every high-end site that buys a server, two or three sites install a free server, depending on their own skill and the support of the network community.
It is difficult to see how the commercial server vendors could displace the large installed base of Apache servers. According to Netcraft, over 350,000 sites are running Apache. Without a compelling feature set that cannot be replicated by Apache's developer community, it seems difficult to imagine a large number of Apache users ever switching.
The other heartening news that can be implied from the survey is the vast popularity of Unix on the Web. Every Apache, NCSA, and CERN site is running Unix, of course, and some amount of the Netscape servers run on Unix as well. It is not difficult to see that perhaps two-thirds of the world's Web sites, and probably more, run on Unix.
I'd be interested to hear from anyone who runs my browser scripts on their agent log, especially if your results differ dramatically from mine. As always, your thoughts and comments on any Web-related topic are welcome. Next month, we'll look at managing Web robots.
About the author
Chuck Musciano has been running Melmac and the HTML Guru Home Page since early 1994, serving up HTML tips and tricks to hundreds of thousands of visitors each month. He's been a beta-tester and contributor to the NCSA httpd project and speaks regularly on the Internet, World Wide Web, and related topics. His book, HTML: The Definitive Guide, is currently available from O'Reilly and Associates. Reach him at chuck email@example.com
If you have technical problems with this magazine, contact firstname.lastname@example.org