Blogosphere 1998: Analysis

The blogosphere arose on 24 April 1998. On that day Steve Bogart made the announcement that he was adopting Jorn Barger‘s recent practice of attributing the source of the links he was posting to his weblog. Barger’s innovation of crediting the origin of his outbound pointers, adopted by Bogart first and by many others afterwards, infused an earlier, mostly dormant network with a dynamic it had previously lacked, turning the aspiration of a self-aware social medium into the reality of such a medium.

What follows is an attempt to substantiate this claim empirically, using a data set of links gathered from online archives. The data set and its limitations are discussed in a separate commentary, which includes call for help stopping the remaining gaps.

Bringing a network alive

The blogosphere was preceded by the News Page Network, a distributed cluster of independent, human-edited news aggregators maintained using UserLand‘s NewsPage software. Each of those news pages was run by a editor who would point out and and lightly annotate news items from across the Web that were of particular interest; each news page featured date-stamped entries that were presented in reverse-chronological order on the front page and sorted into persistent archives located somewhere else in the site’s hierarchy.

The News Page Network aspired to be social. Chris Gulker invoked its existence in October 1997 by gathering the dozen news pages he knew into a list in his site’s sidebar, believing that News Pages were about to usher in “a new, and possibly much greater ‘age of letters.'” Consequently, he hoped that the sites in the network would “proliferate […] and cross refer,” an expectation he acted upon by gingerly initiating the practice of link attribution within the network.

Gulker’s aspiration remained unfulfilled, however. The member sites that he had identified failed to acknowledge the network’s existence and, for the most part, simply ignored each other. The lack of interaction was poignantly expressed in one news page editor’s lament on 1 Apr 1998: “I began doing this a year ago. And I have no idea if anyone actually visits this page!” This lament was symptomatic of the whole network, as any sustained interaction within it proved elusive.

In the first few weeks of 1998, Gulker’s network gained two new members, Jorn Barger and Steve Bogart, who, between them, managed to infuse this dormant network with a dynamic it hadn’t known before. Bogart had been labouring in obscurity on his news page News, Pointers & Commentary since Apr 1997. Barger for his part had first learned about UserLand’s software in August 1997 and started his own news page, Robot Wisdom WebLog, on 17 Dec 1997. Shortly after, Barger discovered Gulker’s network on 5 Jan 1998, and got his site added to the News Page Network list on 25 Jan 1998, to be followed by Gulker’s discovery of Bogart’s site on 13 Feb 1998. Barger praised Bogart’s site the same day as “another fine weblog,” drawing a public thank you note from Bogart.

Barger, intrigued by the News Page Network’s promise of interaction amongst its members rather than the reality of any such interaction, anticipated that the network was going to grow dramatically, predicting in 1997 that “there’ll be hundreds of people maintaining pages like this.” He was convinced that the network needed to serve the purpose of salience generation, of allocating attention to the best that the Web had to offer, which in his analysis required continuous review and selective re-posting of the links offered by other network members. Using the jargon of ‘zines,’ an early form of independent Web publishing, he expressed this idea in January 1998, shortly after finding Gulker’s network, in the alt.ezines newsgroup:

If I were to boil my message down to one point, it would be that web magazines ought to be completely unconcerned about *hosting* content […]. This insight is urgently important for all web authors, imho, and for any surfers who might start such a zine. When it becomes popular, new URLs will propagate thousands of times more efficiently, because each zine author can re-filter a dozen other zines that match their interests.

Accordingly, Barger then set about sampling and remixing the offerings of the sites in the News Page Network, implementing a link attribution scheme on Robot Wisdom WebLog to suit the purpose. On 16 Feb 1998 he took up Gulker’s practice of listing network members, but connected it with his post copy by adding a citation key every time he propagated a link gleaned from another news page. He was meticulous about his attributions and, in addition to offering a prodigious amount of fresh links that hadn’t been posted elsewhere, he racked up a significant quantity of credited links in short order: By 23 April he had posted forty-eight of them, sixteen of which from Brett Glass‘s Your Mileage May Vary (which he may have learned about from Bogart’s recent recommendation), the other thirty-two from members of the NewsPage Network, seven of which came from Steve Bogart.

On 24 April 1998 Bogart decided to adopt Barger’s scheme, vowing to attribute borrowed links in the future. His own sources list, whose members he endearingly referred to as a “merry band of linkers,” initially included Your Mileage May Vary, Robot Wisdom and The Obvious (see bottom of page here). In the remainder of the year, Barger and Bogart each went on to credit another six links to the other’s man’s site.

The new weblogs that came along after this first show of reciprocation did, in Gulker’s phrase, proliferate and cross-link, using both credit links and mention links in unprecedented numbers. Raphael Carter’s Honey Guide Web Log and Avram Grumer’s Pigs and Fishes Weblog both embraced link attributions and generated copious amounts of them. Joe Wilson’s Untitled Weblog happily broadcast its descent from “the tradition of Jorn Barger and the Honeyguide Weblog”. Bill Humphries of Whump Weblog made his entry by recommending Honeyguide Weblog and Robot Wisdom, Brad Graham of BradLands took up link crediting with the discovery of the Pigs and Fishes Weblog, which he attributed to Robot Wisdom, and Michal Wallace of Manifestation soon worried about becoming a “meta-weblog” because of his borrowings from Robot Wisdom. Gulker’s once-dormant network was humming now.

The following line graph tracks three types of intra-network links across the year 1998 and makes clear that once Barger had introduced credit links, they quickly became and remained the most widely used link type from one weblog to another:

line graph: three types of links
Figure 1. Three types of intra-network links across 1998

Noteworthy is also that Robot Wisdom’s number of credit links per month exceeded the aggregate number of the rest of the network at almost any time throughout the year:

line graph: credit links by per month

Figure 2. Credit links per month

The first meme to wend its way through a significant portion of this nascent network, the announcement in June 1998 of a crowd-sourced competitor to Yahoo’s web directory – GnuHoo, which became NewHoo, which became the Open Directory – shows the network doing what it was supposed to do. Slashdot broached the story of Yahoo’s new competitor on 13 June 1998, the Untitled Weblog then reposted the link without attribution on 15 June, as did Pigs and Fishes and Whump, both on 16 June. Honeyguide Weblog then promoted the link on 19 June 1998, crediting the Untitled Weblog, and Manifestation re-promoted it on 20 June, crediting Honeyguide Weblog for it. Link attributions had become an essential part of the network pursuing its core business of salience generation.

Network Measures and Visualisations

Is there a metric that best describes this network? One fairly standard network measure is degree centrality, which determines the number of adjacent links, meaning the number of connections a node has to other nodes:

Table 1
Network Sites and Their Degree Centrality

Robot Wisdom 23 Tomalak’s Realm 6
Scripting News 21 Hack the Planet 5
CamWorld 18 Pigs and Fishes 5
Manifestation 14 Digital Prairie 5
Slashdot 13 Rasterweb 5
Steve Bogart 12 Arts & Letters Daily 4
Chris Gulker 10 Frontier News 4
The Obvious 10 Drudge 4
Whump 10 Memepool 4
Honeyguide 9 Genehack 3
BradLands 9 Need to Know 3
Windowseat 7 Ragged Castle 3
Bump 7 Your Mileage May Vary 3
Flutterby 7 Phil Suh 2
Infosift 7 Untitled Weblog 2
PeterMe 6 MacRonin 2
Obscure Store 6 CopperSky 2
Bud 2


These numbers are used in the following network diagram, which increases a node’s size and moves it closer to the centre as its degree centrality goes up:

network diagram: degree centrality

Figure 3. Degree centrality in network diagram, detail. [Full: Java applet]

Note that the Java applet will highlight a clicked node in conjunction with all of its adjacent nodes, thus nicely illustrating a node’s degree centrality.

Degree centrality, often associated with notions such as the “power” a node is said to hold within a network, may not be the most descriptive network measure when it comes to the early blogosphere. With a nascent network, the process and dynamics of its coalescence are of greater interest than a static construct such as power. Under the common-sense assumption that a network gets built, consolidated and extended iteratively through the allocation of attention from node to node, it stands to reason that the most constructive and supportive node is the one that spreads the most attention around: it’s outlinks to other nodes in the network that raise the tide and lift everybody’s boat a little higher. The simple but infrequently used measure of a node’s outlink count, then, becomes the most descriptive metric. Here are the bare numbers:

Table 2
Outlink Count and Percentage of Overall Outlinks

Robot Wisdom 307 45.4% Rasterweb 5 0.7%
Manifestation 65 9.6% The Obvious 4 0.6%
Steve Bogart 44 6.5% Ragged Castle 4 0.6%
Flutterby 44 6.5% Infosift 4 0.6%
CamWorld 36 5.3% Slashdot 3 0.4%
Whump 22 3.3% Offhand Remarks 2 0.3%
Pigs and Fishes 21 3.1% MacRonin 2 0.3%
Honeyguide 21 3.1% Untitled Weblog 2 0.3%
Chris Gulker 18 2.7% Need to Know 1 0.1%
BradLands 14 2.1% Obscure Store 1 0.1%
Scripting News 12 1.8% Bud 1 0.1%
Windowseat 9 1.3% Digital Prairie 1 0.1%
Hack the Planet 9 1.3% Memepool 0
Bump 7 1.0% Arts & Letters Daily 0
PeterMe 6 0.9% Your Mileage May Vary 0
Frontier News 6 0.9% Phil Suh 0
Genehack 5 0.7% Drudge 0
Psyberspace 0


These data underlie the following bubble chart, in which a node’s size is proportional to its outlink count:

bubble chart: outlink counts for network nodes
Figure 4. Network nodes sized for outlink count, detail. [Full: PDF]

The chart can be enriched by adding all links from the data set, resulting in a network diagram:

network diagram, detail: Nodes sizeds for outlink count
Figure 5. Network diagram, detail: Nodes sized for outlink count, three types of links [Full: PDF]

The full diagram [PDF] may be a bit unwieldy to look at on a computer screen, but its reliance on outlinks rather than degree centrality should render the State of the Blogosphere in 1998 rather well.

An Attention Economy

So Jorn Barger was far and away the most prolific provider of intra-network links throughout the year 1998, and he originally introduced and popularised the most prevalent type of link within the network, the credit link. The fact that Barger’s innovation preceded a conspicuous increase of cross-linking within this network (see figure 1) strongly suggests that the innovation also helped precipitate this phenomenon.

Link attribution, central to Barger’s project of turning the News Page Network into a collaborative filter, allocated attention not only to worthwhile items all across the Web, within the nascent network it also allocated attention to the peers who would find such items. The credit link thereby added an economic incentive to the network: as a micro-payment in the Internet’s true coin of the realm, attention, it rewarded excellence in the business of finding the good stuff among the Web’s overwhelming piles of mediocrity. Link for credited link, Barger and his followers turned Gulker’s inert, notional network into a bustling attention economy, raising the blogosphere in the process.



Blogosphere 1998: Data Set · Commentary · Comment Thread