Joe's "State of the Web sites" Address, Part I
Terry Calhoun, IT Trends Commentator
Society for College and University Planning (SCUP)
University of Michigan
Last summer, some folks at the University of Oregon conducted a comparative
study of 127 university Web sites to analyze technical and design trends. You
might think that the results of such a study would not be appropriate for an
"opinion" piece, but you would be wrong - especially if the report is written
by J'e St Sauver. J
His opinions come through loud and clear, and when I briefly browsed the
report, I thought it would be not only of interest, but with J'e's commentary,
count as an opinion piece, too. Part I, in this issue of IT Trends, covers methodology,
Web server software used, natural minimum Web page size, and the trend toward
segmenting Web sites into "domains" for various user types, such as
prospective students, alums, staff, etc. The first half of J'e's report covers
these areas and it follows. Enjoy!
------------------
J'e St. Sauver
University of Oregon Computing Center
Study Sample
The schools selected for study consisted of the set of all AAU universities
(http://www.aau.edu/aau/members.html),
all Tier 2 or better national doctoral universities from the 2002 US News and
World Report university rankings, and a small number of other colleges and universities
which were traditional comparators or otherwise locally nominated for inclusion.
A study sample dominated by national doctoral universities may do a poor job
of capturing Web trends associated with other higher education segments, such
as four-year liberal arts colleges or two-year community colleges (although
there may be more commonality across those segments than you might think). Virtually
all Web sites are in a state of constant change, so what we saw at a given site
six months ago may not bear any resemblance to what's there now.
In spite of those stipulated limitations, we believe that many sites have tremendous
curiosity about what their peers are doing online, and we hope that the results
of this study will at least provide a foundation for productive local discussions.
Web Server Software
As part of our study, we "fingerprinted" the primary Web server running
at each study site using Internet Periscope
( http://www.lokbox.net/InternetPeriscope.asp
), curious to see what software the study schools were running. (Choice of Web
server software can materially affect the sort of Web site features Web designers
can deploy, as well as the security and performance of the site, etc.). Our
172 study sites, 70.3 percent were running Apache, 12.2 percent were running
Microsoft IIS, and 11 percent were running Netscape Enterprise, with a handful
of other less popular Web servers (WebStar, Lotus Domino, OSU Web Server and
IBM HTTP Server) also being seen. Most large universities ran Apache for their
Web server, with Microsoft IIS generally being associated with smaller or religiously
affiliated schools.
Apache Modules
Apache was designed from the ground up to be highly extensible and highly configurable
via "modules." However, the only Apache modules seen at six or more
study sites were PHP, mod_ssl, and mod_perl. Thus, despite the tremendous number
of ways that Apache can be customized, most sites run with a very conservative
"vanilla" set of extensions, if they use any Apache extensions at
all.
A noteworthy difference between our study sites and general global practice
is that while, by various accounts, roughly 20-22 percent of all global sites
use Microsoft FrontPage extensions, only two sites (1.6 percent) in our study
were running FrontPage extensions. (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 seen at one to five study sites were mod_fastcgi, mod_pubcookie,
DAV, mod_python, mod_layout, mod_auth_pam, mod_ldap_userdir, mod_macro, mod_jk,
ApacheJserv and mod_gzip. See http://modules.apache.org/
for more information about Apache modules.
Natural Minimum Web Page Size
We were also curious about natural minimum Web page size, or the smallest size
window in which a university's Web site could be displayed without causing a
horizontal scroll bar to appear. Deciding on the correct width for a Web page
can be tricky these days since there are huge differences in the screen sizes
people use, ranging 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 can be difficult to design a simple, attractive and easy-to-maintain fixed
width or resizable Web page that works well across that full range of resolutions.
In our study, the median (50th percentile) horizontal page size was 727.5 pixels,
with an X dimension range that went from 486 pixels to 1229 pixels. Vertically
the median was 717.5 pixels, with a range from 409 pixels to 2516 pixels (this
exceedingly "tall" homepage was at a Web sites where the site stacked
badly as the width of the Web page was reduced to its minimum natural width
of 625 pixels). For comparison, Google's home page has a natural width of only
571 pixels.
At least some university Web sites are obviously being quite aggressive about
their use of screen real estate, assuming at least an 800x600 screen resolution
or a willingness on the part of their users to "pan" to see parts
of their Web site.
Web Page Design Trends: Segmentation
We had received some feedback that a growing number of college and university
Web sites apparently had begun to explicitly segment their audience into narrow
categories on their home page. In evaluating that claim, we found that indeed,
78.4 percent (135 out of the 172 sites in our study), now do this.
What is segmentation? A university might choose to offer a number of parallel
versions of its home page: one version tailored for prospective students, another
for current students, a third for faculty and staff, a fourth for sports fans,
a fifth for alumni, a sixth for parents and families of students, a seventh
for donors, an eighth version for members of the news media, and so forth.
Visitors
to the Web sites are typically presented with a choice of Web sites "views,"
often via a link explicitly titled, "Information for..." The site
visitor's page choice is sometimes remembered via a session cookie (or a persistent
cookie). At other times, the visitor's choice is not saved, and merely serves
to determine the next page shown. (Saving audience state via persistent cookies
can lead to bizarre usability problems when done on lab computers or other shared
systems employed by diverse audiences.)
Sometimes the decision to segment reflects a desire to accommodate more direct
links than could reasonably be crammed into a single integrated page; segmentation
obviously makes more real estate available, and reduces the need for a Web page
designer or a Web oversight committee to tell some folks no, their pages won't
be getting directly linked. In other cases, segmentation appears to have been
done simply because other sites were doing it, with no discernible substantive
rationale and largely undifferentiated page versions for each of the "different"
audiences!
That lack of differentiation at some segmented sites is not surprising: it
can be quite difficult to keep half a dozen or more different home pages in
sync and updated. For example, consider the relatively straightforward issues
associated with creating just three or four fresh news items that are uniquely
relevant to each "carved-off" audience! That is quite a task, and
thus it isn't surprising that sites appear to have quickly begun using common
news items across all their "different" segments.
So what about the Web sites that don't segment? The 37 university Web sites
we studied that don't explicitly segment include the University of Oregon itself,
Berkeley, Cal Tech, Cornell, Dartmouth, Harvard, Washington, and Yale. The approach
the current UO home page takes is typical of this group. Rather than explicitly
asking users to categorize themselves, we take a more subtle approach toward
guiding users where they need to go. For example, if you look at the UO's current
home page, you'll notice that we carve off some audience segments via direct
links (Alumni, Sports, The Arts, Visitors), while offering other links that
are important to particular audience segments directly from the home page.
We believe this sort of hybrid design, as used at the UO and a number of Ivy
League institutions and regional competitors, will eventually be seen in virtually
all university home pages because it improves the transparency and navigability
of the site, and because it reduces the user's click count when trying to access
crucial data.
-----------------
So, no real surprises. Most people use Apache, but few get creative with
extensions; our screens are getting larger; and we're segmenting, either at
the top level or one click down, to reduce user clicks. Next
week, we get to portals. How many institutions really have one? Is it a
trend that's already seen its peak?