Electronic publishing through the World Wide Web offers tantalizing opportunities for small-scale operators such as individuals in academic or other non-profit institutions trying to reach a wide audience. Early users of the Web quickly recognized it as a ground-breaking medium for electronic publications. [1] By making it easy to display and read texts online, the Web became a platform for materials that were too specialized, too ephemeral or too experimental for publication as traditional books or articles. However, the recent explosive growth and widespread commercialization of the Web have eroded or at least marginalized small-scale electronic publications. Successful small-scale Web publishing is still possible, but that success must be preceded by careful planning and goal-setting.

I published a World Wide Web site about Balkan history in 1996 to explore the publisher's perspective on the use of online text, a role reversal because as a reference librarian I am most often a consumer of text. Traffic on the site grew dramatically during the Kosovo crisis of early 1999. This paper is a report and meditation on the reasons behind the increase in activity, what exactly took place, and what the experience taught about the online relationship between publishers and information seekers.

Increased public interest in Balkan history is easy to explain: For several months, Western government officials and news media highlighted Yugoslav politics, the plight of refugees, reports of atrocities, and the NATO air campaign against the Milosevic regime. More interesting is a subsidiary question: Why (and how) did the foreseeable growth of public interest in Balkan affairs lead to increased interest in this particular Web site? Answers to that question touch on five areas:

1. The challenge of making Web sites visible to potential viewers;
2. Factors that convert Web site visibility into increased traffic;
3. The potential for interactive communication between authors or publishers and their online readers;
4. The tendency for information seeking to behave as a closed system online; and
5. The interests shared by publishers, authors, and readers in the integrity and authority of online texts.

The Web Site and Its Content

The Web site, Twenty-Five Lectures on Modern Balkan History, is available on a Michigan State University server. It contains the full texts of twenty-five lectures prepared for an introductory course on modern Balkan history that I taught at Swarthmore College in 1995 (I have a background in historical study). A Table of Contents acts as the site's index page, a preface explains why the lectures are available on the Web, and another page states conditions for using the text. I hold the copyright to the content: The text is not in the public domain but is "publicly accessible," a concept developed by David Seaman of the University of Virginia Electronic Text Center. [2] Each lecture is about 4,000 words; at 100,000 words, the total work is about the size of a paperback book, and requires some 700 KB of server space.

It was simple to write and publish the text using commonly available software. The original lectures were written in MacWrite II. They were converted into HTML documents using Word Perfect 3.5 and HTML Editor 1.0 software for Macintosh, later with AOLPress 1.2 for PC (with help from library-systems staff during a Mac-to-PC migration). Winsock 1.1 ftp software handles uploading and modifying the texts in server space provided to library faculty by Michigan State University. All the files were marked up and on the Web by December 1997.

The lectures had not been written with publication in mind. The content is introductory and assumes that readers will turn to other sources for supplementary material. I generalized and simplified substantially for the sake of economy and interest. I cited no sources for statistics or other information except a few direct quotations, and the text has neither footnotes nor bibliography. A book with those limitations would not attract the interest of print publishers, because it would not be competitive with other books in print addressing recent Balkan problems.

As a Web site, however, the same content commands greater interest, for two reasons. With no need to make a profit or recover costs (even for server space, which is university-supported), there is no need to charge for the content. While many books on the Balkans are for sale in the world of print and paper, few are available free. While many books on the Balkans are available in bookstores or libraries, few are available online. Because it is free and available through any networked computer, a text that would be a weak entry in bookstores can be a much stronger entry on the Internet. One might compare the text to a weed: successful in a marginal niche that shields it from overpowering competitors. For better or worse, the nature of the Web creates such niches, places where small-scale publishers and adventurous readers can meet under conditions very different from those defining and limiting mainstream media, for-profit book stores, and traditional libraries. In the world of for-profit publishing, cost-effectiveness is a harsh gatekeeper and takes precedence over quality. However, a text that could never be published by a reputable house can have its day in the sun on the Web.

Visible and Invisible Web Sites

Lesson number one: Putting text on the Web is easy, but bringing text to the attention of readers is hard.

Initial interest in the Balkan lectures site was low, in part because few readers knew that it existed. The number-one obstacle in Web publishing and Web commerce is not placing material online but bringing it to the attention of viewers. No reader can view a site until she realizes it exists. At present, Web surfers rely on two devices to find and reach Web pages: search engines and portals. Web publishers must accommodate the habits of potential readers, while realizing that both devices have limitations based on fundamental characteristics of the online world. For small-scale sites like the Balkan lectures, those characteristics create specific obstacles to wide-spread visibility.

On first inspection, search engines such as AltaVista or Northern Light seem to be promising tools for presenting online material because they are meant to find and list pages from all over the Web. In theory, the right key-word search can bring an interested Web surfer to any Web site. In practice, however, the Web has become too large and complex to make such simple strategies successful.

How big is the Web, and how thoroughly is it indexed by search engines? In June 1999, OCLC's Web Characterization Project estimated that there were over four million Web sites. That figure had increased by 71 percent in one year and by 211 percent in two years. [3] By growing so fast, the Web confounds attempts to describe it. In the latest of several analyses of major Web-search engines, Steve Lawrence and C. Lee Giles reported that no search engine indexed more than 16 percent of the Web, and that new Web pages could remain unnoticed by those devices for months. [4] They suggest that the cost of maintaining large indexes, problems of scalability and bandwidth, and trade-offs between users' wishes for completeness versus retrieval impose limits on search-engine performance. Continued growth of the Web suggests that those problems will increase, and that search engines will remain imperfect systems for connecting Web publications with readers.

Portals are an alternative device for searching the Web, abandoning impossible dreams of completeness in favor of organized lists of Web sites selected to anticipate the interests of Web surfers. Since its origin as a super-bookmark system and "a fun way to collect our favorite sites" [5], Yahoo! has relied on human decision-making instead of automatic Web-crawling spiders to present selected choices systematically. Portal design overlaps with some of the functions of search engines, but the prominence of options such as "Shopping" on a large portal like the Go Network suggests that the site differs in purpose and audience from one like AltaVista that emphasizes elaborate searching capabilities as part of its appeal.

Portals, like search engines, are limited by their nature. Leslie Walker recently described portals as "launching pad[s] for the Web," airport "scheduling billboards," and "electronic channel guides," [6] but in practice portals find it difficult to define their role as clearly as those comparisons imply. Walker notes the contradiction between portals' original purpose — directing users to far-flung parts of the Web — and the current aspirations of Web entrepreneurs to domesticate Web surfers, confine their Web activities (especially shopping) to specific sites, and ultimately to profit from audience loyalty. To that end, "vertical" portals offer a bewildering spectrum of enticing services: but when a portal becomes bewildering, it loses its attractive power.

Susan Kuchinskas points to the same limiting contradiction and foresees the failure of all but a handful of large-scale portals, as Web users come to prefer smaller portals aimed at narrower segments of the market. "The Internet portal business appears to be returning to its beginning as a location to publish content about a chosen niche. In the hunt for a new business mode, it appears as though the Internet will come back to its beginnings as a place to debut content about the niche of one's choosing." [7]

Kuchinskas's comment recalls early expectations about the Web, and implies continued opportunities for small-scale, specialized productions like the Balkan lecture site. However, the failure of large-scale portals to hold the attention of Web audiences suggests that small-scale producers also will fail to reach as many viewers as they would like. If corporate Web-site producers with vast technical and financial resources are unable to overcome the problems of scale described by Lawrence and Giles, small-scale Web publishers should not expect to succeed — unless the terms of "success" are carefully defined. Can Web publishers influence their own fate? Small-scale operators control a limited number of variables.

First, Web-page designers can use markup languages properly when preparing their texts. Correct use of HTML tags like TITLE will maximize the clarity of retrievals. (The TITLE is the phrase that shows up on a search engine's "hit list.") To help users rapidly recognize a site's purpose and value, designers can choose titles, names, and other visible elements carefully. The lengthy descriptive title of the Balkan history web site — "Twenty-Five Lectures on Modern Balkan History" — intentionally includes significant key words for retrieval by search engines, while at the same time identifying the content quickly for anyone who reads the title. The titles of individual lectures perform a similar function, and many search engines can even exploit key words in the lecture texts themselves.

Second, site design can accommodate the widest possible range of viewers. The Balkan lecture site was designed to load quickly, even for users with low-end modems or bandwidth problems. No images are included except for a few hit counters, and HTML tags are limited to a basic core. Except for the hit counters, the lectures meet World Wide Web Consortium (W3C) Web Content Accessibility Guidelines, as presented by the Bobby validator site maintained by CAST, the Center for Applied Special Technology. Each lecture is a separate page, simplifying markup, editing, and uploading by making individual page files smaller. For identification, each page states its URL, the author's name, an e-mail address, the date on which the page was prepared or modified, and a statement of copyright.

Third, intelligent design can ensure that readers who find any part of a larger site understand that more material is available, what it is, and how to find it. Navigation aids should orient users no matter where they are in the site. When it became clear that many users entered the system of Balkan-lecture pages through the retrieval of specific lectures by search engines, an explanation was added to the link back to the main Table of Contents on each page. Because each page has the URL on it, in the text, readers who print a lecture can still find the source site.

In practice, those decisions are less straightforward than the description implies, because designing a strategy for site navigation requires choosing among competing priorities. On the Balkan lecture site, individual lecture pages load rapidly, but the whole system of lectures is connected through the Table of Contents, making it harder to move from one lecture to another. The design prompts users to go through the lectures in order, but also slows down their progress by requiring repeated visits to an index page. There are obvious alternative approaches, such as a comprehensive linked list of lectures at the top of each page, or buttons on each page for "Previous lecture" or "Next lecture." Either of those approaches would deemphasize the Table of Contents and its role as a summary and an orientation device.

Those techniques primarily address search-engine users. Promoting a site via structured portals is a different matter. Some portals invite visitors to nominate Web sites for consideration. I submitted my site's URL and a description to several portals, [8] with no apparent result until the Kosovo crisis attracted serious attention. In October 1998, The Mining Company (which has since merged with a larger portal, About.com) sought permission for links to some lectures. In early 1999 the site simply appeared on Yahoo! without notice, and is now available through many portals.

The lecture site also appeared on other, less influential, small-scale Web sites. In the Spring of 1998, the site became a free offering of the ConnecText Catalog after an interesting online discussion with its creator, Professor Lynn D. Nelson of Virginia Commonwealth University. After the crisis began, the lecture site was linked to several resource lists developed for various university classes, international think tanks, and relief organizations with an interest in Kosovo.

Observations by Lawrence and Giles in July 1999 suggest one advantage enjoyed by small-scale publishers, especially those in nonprofit environments. Lawrence and Giles reported that an average of fifty-seven days passed between the creation of new documents on the Web and their discovery by search engines. In other words, search engines might not have listed a new Kosovo-related Web page created at the beginning of the NATO air campaign until the crisis was almost over. The relationship between advertising, income, and the cost of maintaining little-used Web pages suggests that few for-profit Web publishers or portals would have supported Balkan-related pages before the crisis heightened public interest in the region. On the other hand, Web pages supported by academic institutions carry no burden of profit, yielding this potential advantage: Such sites can be established well in advance of reader interest, persist despite low traffic levels, and undergo indexing through the slow but relentless activity of Web spiders. The Balkan lecture site was available for more than a year before the crisis, and such automated indexing did take place.

Some large portals can offset that tendency toward invisibility for new Web pages through previously established prominence. If a potential viewer identifies a topic with "the news," she is likely to seek out a known news portal such as CNN.com, where readers can be directed to brand-new pages. Name recognition of that kind is precious, because it sidesteps the defects in search engines and portals. CNN claims that "Traffic has grown to 100 million page views per week; that number can spike substantially when major news events occur." [9] Some booksellers also are able to tap name recognition. After the sudden death of John F. Kennedy, Jr., in July 1999, sales of books about the Kennedys soared at both virtual and traditional "portals" selling books. Amazon.com and Barnes & Noble chain stores alike recorded strong sales. [10]

Hot Web Sites and the "Nova Effect"

Lesson number two: Sustaining interest in a Web site can be as difficult as creating that interest in the first place.

When a star goes "nova," it can flash from obscurity to brilliance in a short period of time, but that star will later resume its dim existence. The same thing can happen to Web sites, as illustrated by changes in traffic on the Balkan lecture Web site. If success is measured by prominence alone, success can be transitory and hard to sustain.

A Web site can become "hot" for reasons beyond the control of its creators. The Kosovo crisis clearly led to expanded site traffic on my Balkan-lectures site. Vietnam has been called America's first "television war" and anyone who watched the live CNN broadcasts of the Baghdad skyline might consider the Gulf War as the first "cable war." The NATO air campaign against Serbia may be remembered as the first "Internet war." [11] News filtered out of beleaguered Kosovo by e-mail, anti-NATO hackers meddled with United States government Web sites, and the warring parties competed for world sympathy through the World Wide Web as well as other media.

When the crisis led to more e-mail contacts, I asked the Michigan State University Libraries systems staff to provide simple hit counters, which I added to selected pages on the site: the Table of Contents, the preface, and three sample lectures. Although counters failed and reset to zero several times, by regularly recording hit levels offline it was possible to keep a plausible count of cumulative hits. This was another aspect of the site affected, and weakened, by an absence of thorough planning. Not having anticipated the sudden growth in traffic due to the Kosovo crisis, I responded hastily to events; not having spent time explaining the site and its purposes to systems staff, I failed to ask for the best support they could have offered. In retrospect, using software such as WebTrends would have yielded far more interesting data.

The number of hits on the site rose and fell with the urgency of the crisis. [12] Web traffic peaked during the week of April 21 to 27, when NATO aircraft attacked major buildings in downtown Belgrade for the first time. During the same week, spring-term research papers were due in schools across the United States. As a result, three times as many readers hit the Table of Contents page that week than during the week before. Thereafter, activity fell off steadily, except during two major events: NATO's accidental attack on a refugee convoy, and the exodus of refugees back into Kosovo after the cease-fire.

How "successful" was the Balkan lecture site? The answer depends on one's measure of success. At its peak, the Table of Contents page registered 3,100 hits in a week. Compared to the volume of hits on a site like CNN, that figure is pitifully low; compared to past volume on the Balkan lecture site, it was quite high. High traffic levels can be hard to sustain. After peaking at 3,118 hits per week, traffic on the Table of Contents page fell to 1,323 hits, 756 hits, 833 hits, 789 hits and then 384 hits in the next five weeks. No subsequent week saw more than 500 hits. Short of a revived conflict, there is no reason to expect a dramatic reversal in those numbers. Pre-crisis traffic levels are unknown, but post-crisis figures remain close to 200 hits per week, barely 6 percent of the peak number.

If traffic figures are going to be the measure of success, the means for collecting them should be considered in the early stages of site design. Navigation schemes will influence the result. The Table of Contents page on the lecture site typically registers three times as many hits as the most popular individual lecture, and ten times as many hits as less popular pages. However, the site's navigation system and simple counters are not really adequate for analyzing movement within the site. Repetitive hits on the Table of Contents by diligent readers retrieving a series of lectures can be hard to distinguish from brief, casual inspections by Web surfers who glance at the Table of Contents and then leave the site. And a hit on a lecture page is not evidence that a lecture has been read. Small-scale Web site publishers need to weigh the significance of the numbers they try to collect, against the time needed to choose and manage a software package that allows traffic analysis.

Web publishers defining success as traffic volume also face hard choices if their site completes the "nova" cycle — that is, if traffic reverts to low levels. Do future prospects suggest keeping the site intact and waiting for lightning to strike again? Can the site evolve to follow the changing interests of former visitors? Is it time to dismantle the site? Is the site still valuable to its original audience, despite the drop in traffic? For nonprofit operators, of course, those questions assume less urgency.

The "Iceberg Effect" and the Expectations of Site Visitors

Lesson number three: Online readers may expect more from a Web site than the author or publisher is in a position to provide.

E-mail provides both intentional and accidental information about people who visit a Web site. The Web is so new, and has so many new users, that many Web surfers will misunderstand the purpose and scope of online information and surprise, baffle, or impose on Web authors and publishers.

First let it be said that e-mail traffic figures confirm the "nova effect" in which a Web site rapidly becomes "hot" and then "cold" again. That was the case with the Balkan-lecture site. In the nine months between putting up the site and the beginning of international interest in Kosovo (January to September, 1998), I received messages about the lectures from only eight individuals, about one message per month. As NATO increased its pressure on the Yugoslav government short of war (from October 1998 through late March 1999), twenty-eight messages arrived from twenty-one individuals, roughly one message per week. During the NATO air campaign (March 24 to June 10, 1999), I received ninety-three messages from thirty-two individuals, about one message per day. Finally, after the signing of peace terms, traffic reverted to pre-war levels: seventeen messages from twelve individuals in twelve weeks. It is worth noting that no one has ever responded to the site by surface mail, even though a postal address is provided in addition to the "mailto:" tag.

Besides affirming the "nova effect," those messages revealed some motives behind online information-seeking. Of the seventy-three individual correspondents, return addresses suggested that fifty were in the United States. Based on surnames and self-identification, almost half of that group had ethnic roots in the Balkan region. Eight people wrote from other English-speaking countries: the United Kingdom, Australia, Ireland, and Canada. Six contacts came directly from southeastern Europe: Greece, Yugoslavia (Serbia), Croatia, Hungary, and Turkey. The other nine messages came from eight developed countries: Spain, the Netherlands, Belgium, Germany, Denmark, Poland, Israel, and Japan. Three quarters of the globe was silent: Africa, Latin America, and Asia outside Israel and Japan.

Some patterns emerge in readers' motivation for using an online source like the Balkan-lecture site. The Web site was helpful for English-language speakers living abroad who had limited access to books in English through local libraries or bookstores. It was consulted by student users who routinely turn to the Web for some or all of their research needs, whether or not they use traditional print tools.

E-mail messages varied widely in tone, in part because the site provided no guidance beyond an invitation to correspond with the author. As a result, I not only received a large amount of e-mail during the crisis, but also e-mail with challenging content: questions about further reading, genealogical research, or translating texts from Balkan languages into English. There were political diatribes, and concise, sometimes blunt, critiques of the site and its purposes. Of those capsule reviews, the most favorable was "Your info wuz da bombs" from a Canadian high school student; the least favorable, a pithy "Shame on you" from a reader with unstated but contrasting views. [13]

Readers' expectations about online communication with authors are diverse and intriguing. I acknowledged almost every message. After receiving a brief note, one correspondent wrote back to say, "Thank you for your response (and speedy too!). I didn't think that an academic would even bother giving a response, let alone a dignified one." [14] Another accused me of being a paid NATO agent. Most comments were laudatory and intelligent, however.

Guidelines, stated or not, can help manage torrents of e-mail if a Web site becomes "successful," and reduce the amount of time required to respond. Forty-eight of the seventy-three people who responded to the site wrote only once. Sixteen people sent a second message, usually a simple thank you for help with some question. The other nine correspondents kept on writing back — three, five, ten, eleven, or even twenty-nine times in a few weeks. Extended correspondences tended to have a single theme: arguments about NATO policy, for example, or requests for precise information about recent news. In many cases, I could only remind readers that the Web site dealt with historical, not current, events.

Many readers drew unrealistic conclusions from the lectures. In their minds, the Web site was something like an iceberg. They assumed that the visible portion amounted to perhaps 10 percent of a submerged mass of information. That "iceberg effect" led some readers to ask for facts about medieval or contemporary affairs even though the site is explicitly focused on the period from 1790 to 1989. Despite clear statements that the content was introductory, other readers expected to be given additional information in depth.

Over time I realized that I was most interested in particular user issues, and sought information about them. I amended my preface to the site to ask two questions: "How did you find the site?" and "Why did you find the site useful or interesting?" A few recent readers have responded with relevant comments. Unless Web publishers and authors guide e-mail in some manner, they are unlikely to gain much from a potentially time-consuming chore.

The Internet as a Closed System

Lesson number four: Not everything is available on the Internet, but some Internet users behave as if it were.

Some readers chose to use the Balkan-lecture Web site for practical reasons. Those users included Americans living abroad with little access to English-language publications. Others, however, opted for online research because their preference in media — Web versus paper — was their primary criterion for selecting sources, regardless of the quality of the information. [15] Such users tried to function within the Web as a closed system, ignoring valuable print resources.

Librarians see that students prefer Web tools over paper, but my experiences as a Web publisher uncovered similar impulses in some unexpected constituencies, such as the press. Three news organizations contacted me during the crisis, after discovering the lectures through search engines. All three used the telephone; two used e-mail as a backup. Two contacts came from research units at nationally prominent news organizations looking for experts for on-the-air interviews; the third was a small consulting firm preparing Web-based materials on Kosovo for reporters.

Throughout the crisis, readers pressed me to provide detailed information about current events. Historians are loath to speak in the absence of archival evidence. Instead I referred those readers to other sources, especially major news organizations. It was startling, then, to see an ironic circle completed when news organizations approached me as a source for the very information that I had been assuring readers was best sought from media outlets. In fairness, the press makes extensive use of sources outside the Web: nevertheless, it was interesting to encounter the same impulse to rely on the Web among traditional media staff. The seductive simplicity of Web-based "research" is difficult to resist, especially when deadlines loom — whether for reporters or for students.

Authority and Integrity for Online Texts

Lesson number five: Establishing the authority for Web-based information is important for both producers and consumers.

If researchers prefer online information, the authority and integrity of documents on the Web becomes a key element for truth and accuracy. The most concisely thoughtful e-mail comment from a reader of the Balkan lectures was this: "Your website is fascinating and gives every appearance — I'm no scholar in this field — of being authoritative." [16] That message hits the nail on the head: someone who is "no scholar in this field" may have trouble distinguishing actual validity from the appearance of validity on a polished Web site. When format becomes the principal criterion of choice, and Web-based research functions within a closed system, the road is paved for persuasive but bogus information.

Issues of site authority and textual integrity matter for both users and authors. Users should require statements of authority for materials on the Web. My Balkan-lecture site states my credentials as both a historian and a librarian. The opportunity for e-mail communication is a potentially helpful check for readers; so is access to other Michigan State University Web pages, which allow them to verify my position as member of the library faculty.

Experience with the Balkan-lecture site suggests two related issues of authority for authors and publishers. First, controversial topics such as Balkan politics are ripe for disagreement and misinterpretation. Second, by providing texts in a "publicly accessible" format such as a Web page, authors diminish control over the accuracy of their intellectual output. The more success one has in disseminating a text to the far corners of the globe, the more difficult it becomes to detect and correct degradations of text, loss of context, or deliberate tampering. Someone who copies a document from its original site can easily tamper with the text and post it elsewhere in an altered form. It can be difficult for either readers or authors to recognize, let alone correct, this kind of meddling.

I encountered that situation several years ago, when an online book review I had written for H-Net was altered and disseminated through a Usenet news group. Like many of the Balkan lectures, the text dealt with potentially controversial topics. After receiving several strongly critical e-mail messages from Balkan émigrés living in Australia, it became clear that the word "not" had been removed at a crucial point in the review, before an altered cut-and-paste text was posted to a discussion group. By pointing correspondents to the H-Net online archives, I soothed some feelings, but readers who failed to contact me may never have realized that they were working from a defective text. Of course, the Web makes it possible to contact the author of an online text easily, especially if a text retains the author's e-mail address. In this regard, Web texts have some advantages over paper: if they are easier to alter, they are also easier to correct.

Small-scale Web publishers are more likely to suffer this kind of time-consuming controversy, than actual financial losses. The hacking attacks cited by Taylor and Mellow in their articles on the Balkan cyberwar of 1999 point to more extensive, malicious and damaging tampering in the future, as online information becomes a weapon of war. Episodes of that kind erode the Web's credibility, and with it the Web's ability to level the playing field between large and small publishers.

Copyright is of limited use in such cases. The Balkan-lecture site "Conditions of Use" page addresses some copyright issues, but is a basically passive device. Keeping the present site as the single authoritative source for the text is another remedy. I modified the site at times, correcting minor errors or adding features, and that ongoing control increases the odds that readers will encounter only the latest state of the text. Problems like pirated copies are more serious for commercially valuable materials; for nonprofit sites the issues are intellectual rather than financial.

At a minimum, Web publishers can give readers the information they need to verify a source: the author's name, an e-mail address, the date on which the page was prepared, a URL, and a statement of copyright. In the case of the Balkan-lecture site, the preface also adds some useful context: the origin and purpose of the project; the identity and authority of the author; and some tips about citing Web pages in research papers.

Conclusions

There is every reason to think that the World Wide Web remains, and will remain, a valuable outlet for small-scale publishing. Web authors and publishers in academic and non-profit environments are largely immune from the intense competition for revenue that characterizes for-profit sections of the Web, and small-scale publishers can use the portals and search engines created by giants like Yahoo! and AltaVista to promote some of their efforts. On the other hand, large-scale commercial Web sites throw increasingly large shadows that make it harder for small-scale operators to remain visible.

At the same time, small-scale Web authors and publishers in academic or nonprofit environments are largely immune from the immediate effects of those battles to control the Web. At present, the portals and search engines created by giants like Yahoo! and AltaVista provide platforms that are available for small-scale publishers, however unreliably.

Some observations based on several years' experiences with the Balkan-lectures Web site may help other small-site publishers invest their resources wisely.

• Think carefully about the purpose behind your site before you create it. Apparently minor decisions can strongly influence the balance between positive and negative outcomes.

• Keep in mind that the benefits of a Web site can be measured not only in dollars, but in experience gained, contacts made, institutional and personal visibility, and services provided. At the same time, the costs of maintaining a Web site are not measured in dollars alone; considerations of time are especially important.

• Define the terms of your own success. Be realistic. When will you expect to know if a site has "succeeded" or "failed", and if it "fails," what must you do about it?

• Specialized sites, such as those created by academics, occupy narrow niches. That applies both to subject matter, and time. If your site is of interest to a majority of people in a tiny subdiscipline, is that sufficient? If your site is famous for fifteen minutes, is that a victory or a defeat?

• Have a strategy for relating to your readers. Online interactivity is a potential strength for Web publishing, but take steps to make sure you control interaction before it controls you.

• Pay attention to the integrity of your materials. If accuracy or copyright are compromised, will you pursue some remedy?

With forethought and planning, small-scale Web publishers can still make important contributions to advance their own interests.

Steven W. Sowards is Head of Main Library Reference at the Michigan State University Libraries, East Lansing. In addition to the M.L.S., he holds a Ph.D. in History from Indiana University with a thesis on Habsburg foreign policy in Macedonia in the period preceding the Balkan Wars. Prior to coming to Michigan State in 1996, he was Humanities Librarian at Swarthmore College, where he also taught occasional courses on Eastern European history as an adjunct instructor. His home page may be found at http://staff.lib.msu.edu/sowards/staff/index.htm. You may contact him by e-mail at sowards@msu.edu.

Footnotes:

1. In 1994, for example, James Powell wrote "The World Wide Web system is also attractive due to the potential for the appearance of value-added products encoded in HTML. SGML is an attractive publishing format due to the flexibility of the end-product, and provides new opportunities for publishers." "Adventures with the World Wide Web: Creating a hypertext library information system," Database 17 (1)(February 1994):59+. Available on line for licensed ProQuest Direct subscribers.

2. The University of Virginia Electronic Text Center states its Conditions of Use online. For more about the center, see David Seaman, "'A library and apparatus of every kind': The Electronic Text Center at the University of Virginia," Information Technology and Libraries 13 (1)(March 1994):15+ . Available on line for licensed ProQuest Direct subscribers.

3. Figures from the OCLC Office of Research Web Characterization Project: Statistics page [formerly http://www.oclc.org/oclc/research/projects/webstats/statistics.htm].

4. Steve Lawrence and C. Lee Giles, "Accessibility of information on the web," Nature 400(6740) (8 July 1999): 107-109, available on line for licensed Nature subscribers. Their earlier analyses include "Searching the Web: General and scientific information access," IEEE Communications Magazine 37 (1) (January 1999):116-22; and "Searching the World Wide Web," Science 280(5360)(3 April 1998): 98-100, available on line for licensed Science subscribers.

5. Quoted from a message from our founders, David Filo and Jerry Yang on the Yahoo! home site.

6. Leslie Walker, "Some Light Through the Portal," Washington Post, 19 August 1999, p. E1; available on line for licensed Lexis-Nexis Academic Universe subscribers.

7. Susan Kuchinskas, "Content's Comeback," Adweek 40 (15 March 1999): IQ22, E1; available on line for licensed Lexis-Nexis Academic Universe subscribers.

8. The "How to suggest your site" page at Yahoo! is a representative example.

9. Quoted from CNN's "Advertiser Benefits" page. [Editor's note: this page is no longer available, so the link was removed on 05-31-2001].

10. Hillel Italie, "For Kennedy authors, a time to mourn - and to write," Associated Press State & Local Wire 20 July 1999; available on line for licensed Lexis-Nexis Academic Universe subscribers.

11. As early as 1969, Michael J. Arlen used the phrase Living-Room War as the title of his book about Vietnam. On the "cyber war" or "Web war" that accompanied the NATO air assault, see Philip M. Taylor, "Propaganda and the Web war," World Today 55(6)(June 1999): 10+, available on line for licensed ProQuest Direct subscribers; and Craig Mellow, "Who won the Balkans cyberwar?", Institutional Investor 33(8) (August 1999): 22+, available on line for licensed ProQuest Direct subscribers .

12. It was easy to compare hit traffic with events by consulting online crisis chronologies such as the Washington Post "Kosovo 1980s to 1999" time line.

13. Personal e-mail (15 October 1998 and 18 April 1999). In the interests of their privacy, I see no compelling reason to identify specific correspondents.

14. Personal e-mail (13 May 1999).

15. Sherry Turkle emphasized that point in a lecture at the Michigan State University Libraries in October 1996, after which the author made a more pointed effort to address issues of authority and explanation on the lecture Web site, especially through the Preface. On the influence of self-image for computer use, including "cyborg consciousness" among young people who tend to prefer online research, see Sherry Turkle, "Computational technologies and images of the self," Social Research 3(Fall 1997): 1093-1111, available on line for licensed ProQuest Direct subscribers.

16. Personal e-mail (10 June 1999); my emphasis.

Links from this article:

About.com. (http://about.com/)

AltaVista. (http://www.altavista.com/)

Amazon.com. (http://www.amazon.com/)

Barnes & Noble. (http://www.barnesandnoble.com/)

Bobby. (http://www.cast.org/bobby/)

Center for Applied Special Technology. (http://www.cast.org/)

CNN.com. (http://www.cnn.com/)

CNN: Advertiser benefits page. (http://cnn.com/ads/e.market/adben.html)

ConnecText Catalog. (http://www.connectext.com/)

Go Network. (http://www.go.com/)

H-Net. (http://www2.h-net.msu.edu/)

Lexis-Nexis Academic Universe. (http://web.lexis-nexis.com/universe)

The Mining Company. [formerly http://history1700s.miningco.com/]

Nature. (http://www.nature.com/nature/)

Northern Light. (http://www.northernlight.com/)

OCLC Office of Research Web Characterization Project: Statistics. (http://www.oclc.org/oclc/research/projects/webstats/statistics.htm)

ProQuest Direct. (http://www.bellhowell.infolearning.com/proquest/)

Science. (http://www.sciencemag.org/)

Twenty-Five Lectures on Modern Balkan History. (http://www.lib.msu.edu/sowards/balkan/index.htm)

University of Virginia Electronic Text Center, Conditions of Use. (http://etext.lib.virginia.edu/conditions.html)

Washington Post, "Kosovo 1980s to 1999". (http://washingtonpost.com/wp-srv/inatl/longterm/balkans/timekosovo.htm)

WebTrends (http://www.webtrends.com/default.htm)

Yahoo! (http://www.yahoo.com/)

Yahoo!: Message from our founders. (http://howto.yahoo.com/message/)

Yahoo!: How to suggest your site. (http://docs.yahoo.com/info/suggest/)