How do you map the Web? An obvious approach is to map the hyperlink structures as some kind of graph, either in two or three dimensions. This is quite easy for individual sites, and tools exist for web masters and designers to do this [1], but to map the many millions of web pages and their interweaving hyperlinks is an altogether more challenging task. One could argue that it is also a foolish task as the Web is in fact its own map, and it is impossible to produce a better one. This idea harks back to Jorge Luis Borges' famous Cartography fable, with the Web being both the territory and the map [2]; by mapping it in detail one simply ends up creating it anew.

      While the Web as map and territory is conceptually true, there is still a need for maps of it that simplify and generalise, highlighting the interesting and hiding the superfluous detail. What people want is the map of the Web from a distance, floating above it somehow, so they can get a sense of the broad view of the information landscape; not the highly detailed graph-like mesh of every single page and link.

Links as Citations Citation Indexing In the July 15th 1955 edition of Science, Eugene Garfield published Citation Indexes for Science. He argued that a systematic analysis and indexing of the scientific literature was needed to highlight the most influential publications. Moreover he argued that such a system would help scholars keep track of criticisms of oft-cited publications and thereby reduce the likelihood that inaccurate or misleading data would be naïvely repeated. In the 1990s Eugene Garfield continues to publish–a recent article in The Scientist warns against politicization of citation indexing.[ed.] » read "...about Uncitedness"

      A few years ago Tim Bray took the challenge and produced some fascinating, and very unconventional, maps of the Web landscape [3]. Bray is a programmer and all-around general hacker of text, hypertext and XML, based in Vancouver. Before he could map the Web, he needed to take some measurement of it using a large search engine index as the easiest source. He calculated key metrics on structure of the known Web in 1995, comprising a mere eleven million pages from about 90,000 sites, to answer the fundamental questions - how big it is? how wide is it? where is the centre? how interconnected is it? These questions are still very relevant to academic researchers and commercial pundits today [4]. Much of the research into understanding Web morphology focuses on the analysis of the human-built hyperlink structures, the aim being to improve current Web searching tools and develop new searching algorithms to find the elusive nuggets of gold that lie hidden in the ever growing expanse of the Web. The importance of hyperlinks rests on the fundamental assumption that they are conscious recommendations people make to Web sites or pages that they think are useful or of interest. In this manner hyperlinks play the same role as citations in scientific literature. There are decades of research into citation analysis, pioneered by Eugene Garfield [5] and the Science Citation Index, trying to understand and map out the intellectual structures of scientific ideas and disciplines.

      Bray spent time analysing the hyperlink structures of the Web and found interlinking between sites was surprisingly sparse. Most links were local, within a site and only a few key sites acted as super-connectors tying sites together - Yahoo! being the epitome of this, the master weaver of the Web. He derived two intuitive measures of Web site character based on hyperlinks - visibility and luminosity. Visibility is a measure of incoming hyperlinks, the number of external Web sites that have a link to them. He found a select few highly visible sites with many thousands of incoming links, they are the bright stars of the Web. In 1995, the most visible Web site was that of the University of Illinois, Urbana-Champaign (UIUC), the home of the Mosaic browser [6]. The vast majority of sites had only a few links to them and nearly five percent had no incoming links at all. Measuring the reverse, the outgoing links, determines a site's luminosity. The top few percent of most luminous sites obviously carry a disproportional amount of navigational workload. Yahoo! was the most luminous site in 1995 and probably still is today.

      Bray used statistical characteristics to map the key landmarks of the Web in 1995, highlighting the largest, most visible and connected Web sites. The visual metaphor employed to represent Web sites was radically different from conventional maps, having more in common with architectural models than planar cartography. He used abstract, 3d models which I think look like a cross between the pieces on a chess board and totem poles. (He termed his model ziggurats, the ancient stepped pyramidal temples.) I think the analogy with totem poles, created by native American peoples of Pacific Northwest [7], is particularly apt because they were powerful emblems of a family or clan. Bray's models were totems of Web sites in 1995.

Cyberspace as seen by hyperlink lights. Statistical analysis of Internet hyperlinks formed the dataset underlying the structure of Tim Bray's map of the Internet as it existed in 1995. These flat images were captured from a 3 dimensional mapscape. Each totem pole like structure represents a web site. The size of the basal disks graphically represents the quantity of content on the site, its height denotes the site's visibility on the net measured by a count of hyperlinks coming into the site from other sites and the spherical headpiece represents the site's hyperlinks out to other sites– a quality Bray calls luminosity. [ed.] maps ©Tim Bray reproduced by permission » view maps

      The web site totems visualise the degree of luminosity and visibility of a site, along with the size of the site and the broad category it was in (government, education, commercial, etc.). The basic graphic properties of the totem, size, shape and colour, were used to encode these four dimensions of a web site. The overall height of the pole represented visibility, the width of the pole is size of the site, in number of pages. The size of the globe atop the pole maps the site's luminosity and finally colour coding was employed for different categories (green for university, blue for commercial, red for government agencies). The totems were also labeled with the domain name of the site for identification. The spatial layout of the totems on the ground plane was based on the strength of the hyperlink ties between them.

Mapping family myth Totem poles were carved by native Americans of the Pacific northwest as graphic representations of family history, relationships of a particular homesite to mythic events, and to memorialize deceased ancestors. An interesting collection complete with photographs, interpretations, and something about the family history and lore that went into making them is hosted by the Royal British Columbia Museum. [ed.] » go to the Royal BC

      The two images show a field of totem poles mapping the Web sites at the very core of the Web in 1995. Further from this core region there would be many thousands other, but most would be minuscule in relation to the mighty totems at the heart. The totemic models were constructed automatically from site statistics using VRML. These examples are simply static screen-grabs of the 3d models.

      Today, Bray's web site totem poles stand as historic monuments to the early days of the Web. They have yet to be repeated, but in a recent email, he reported that he is back in the Web mapping business, spending eighty hour weeks working on what to map and appropriate visual metaphors to represent it. I asked him what his 'ultimate' map of the Web would be like, his answer was for a characteristically unconventional visualisation,

"I don't see it like the visions of Gibson and Stephenson - glowing abstract colour-on-black geometries. My vision is more like a bazaar in Damascus or a mall in Kowloon - a lot of disorderliness, organic rather than planned growth, some parts smell bad, full of surprises."

      We should keep an eye on his home page (http://www.textuality.com/) and wait for the sounds and smells of a new disorderly Web map to emerge.