|
|
Joe St Sauver, Ph.D.
Director, User Services and Network Applications
joe@uoregon.edu
At the request of the University of Oregon administration, we conducted a comparative study of 172 university websites this summer.
The schools selected for that study consisted of the set of all AAU universities ( http://www.aau.edu/aau/members.html ), Tier 2 or better national doctoral universities from the 2002 US News and World Report ranking ( http://www.usnews.com/usnews/edu/college/rankings/rankindex_brief.php ), and universities that have been traditional comparators or were otherwise locally nominated as being of special relevance or interest to the UO. We looked at 172 universities in all (for the complete list of universities, see http://darkwing.uoregon.edu/~joe/2003-web-study/sites.txt ).
We summarize some of those results here because we believe they are of general interest to UO web page authors, and because they may be of interest to other universities as they think about their own websites.
Before digging into the results, we should also mention that university websites, like all websites, are frequently changed. The information we present below will obviously drift over time as various universities make incremental changes or transition to entirely new website designs.
I. What Web Servers Are Being Used?
Every website runs on top of some sort of web server software. Choice of web server software is a fundamental one, one which can have a profound influence on factors as diverse as:
Comparing academic web server use to overall web server use: For the last eight years, Netcraft has been tracking global web server software market share (see http://news.netcraft.com/ ). In the latest Netcraft report, (dated August 2003 and based on their evaluation of nearly 43 million websites worldwide), Apache ( http://www.apache.org/ ) was used by roughly 64% of all sites, Microsoft IIS ( http://www.microsoft.com/iis/ ) was used by roughly 24%, and the remaining 12% consisted of a variety of other web server software products. This is quite comparable to the market share reported by Security Space ( http://www.securityspace.com/s_survey/data/200307), which shows Apache at 66.6% and Microsoft at 24%, with the remainder comprising a variety of other web server software products.
In the case of our more focused study of 172 higher education web servers, we "fingerprinted" the primary web server at each of our selected sites using Internet Periscope ( http://www.lokbox.net/internetPeriscope.asp ). Of our 172 selected study sites, 121 sites (70.3%) were running Apache, 21 (12.2%) were running Microsoft IIS, and 19 (11%) were running Netscape Enterprise.
Other less popular web servers included WebStar (running at Northeastern, Reed, and the University of Connecticut), Lotus Domino (running at Seton Hall and the University of the Pacific), OSU Web Server (running at Miami University of Ohio and Ohio University), and the IBM HTTP Server (running at LSU).
A few other sites, presumably motivated by security concerns, had taken steps which made it impossible for us to fingerprint their web server software.
Some observations about the web server software results:
II. Apache Modules in Use
Because Apache is modular and its base functionality can be extended via a variety of modules, we also looked to see what (if any) modules were reported as being used in conjunction with Apache. By looking at the modules deployed, we can tell a variety of things, such as the extent to which a given website is interested in offering dynamic content, or has the ability to offer secure (encrypted) web pages from its primary server.
The most popular add-ons seen at least half a dozen of our study sites were, for the most part, the ones you'd expect to see
(All study site percentages figured on a denominator of 121 Apache-using sites. Comparative data is from Security Space's Apache Module report for July 2003; see http://www.securityspace.com/s_survey/data/man.200307/apachemods.html ).
The one noteworthy difference between modules in use at study sites and global trends is that while 21.92% of all global sites used the FrontPage extensions, we saw only two sites (1.6%) in our study that were running FrontPage. (There have been a variety of security issues associated with the FrontPage extensions which may have resulted in this difference in penetration.)
Other Apache modules or add-ons seen only at a few sites included mod_fastcgi, mod_pubcookie, DAV, mod_python, mod_layout, mod_auth_pam, mod_ldap_userdir, mod_macro, mod_jk, ApacheJServ, and mod_gzip. For information on these or other Apache modules, see http://modules.apache.org/
III. Natural Web Page Size
We live in a time when there's an exceptionally wide range of screen sizes in use: everything from 640x480 (307,200 pixels) on small legacy monitors all the way up to 2048x1536 (3,145,728 pixels) on huge high resolution tubes--a full order of magnitude difference. It's difficult to design a web page that works well across that full range of resolutions.
Some sites attempt to address this issue by making their pages resizable, but doing that in a robust way can greatly complicate the page design process, and will often fail (or produce bizarre results) as users increase or decrease page size beyond relatively modest limits. The page that deserves the "gold star for near-perfect resizability," (although the design has other issues, such as its distracting use of animated page elements, that preclude giving it an unqualified thumbs up) would be New School University's site, http://www.newschool.edu/
Other sites simply create a fixed size page that will work on even the smallest of displays, accepting the risk that users running on large high resolution displays will see a disappointingly small (and hard to read) page; Montana State and the University of Dayton are examples of pages that take this approach. On the other hand, if you elect to optimize your web pages for larger displays, users on smaller tubes may find themselves having to scroll around in order to do even the most basic of tasks, and items which only become visible when the page is scrolled risk not being seen at all.
We were interested in what our study sites had chosen as a "natural minimum page size," or the smallest page size at which a user would not see a horizontal scroll bar appear.
The median (50th percentile) horizontal page size for our studied sites was 727.5 pixels, with an X dimension range that went from 486 pixels (University of Oklahoma) to 1229 pixels (Arizona State).
Vertically, the median page size was 717.5 pixels; the Y range extended from 409 pixels (Utah State) to 2516 pixels (the University of Kansas website, a page which stacked badly when the width of the web page was reduced down to its natural width of 625 pixels).
For comparison, the UO's natural page size is 603x694. This natural width is substantially smaller horizontally than the study sites median width, but is right in the same ball park as Google’s natural width of 571 pixels. Our height is quite close to the study median.
The images of the study websites from which these measurements were taken are available online at http://darkwing.uoregon.edu/~joe/new-web/
Next time…In this issue of Computing News, we talked about some of the mechanical issues associated with university web page delivery. In the next issue, we'll describe the trends we're seeing with respect to the actual design of higher education home pages.
Some of the university web page trends we've analyzed and will be reporting on next include:
Stay tuned for more on the evolution of university web sites!