`
`Proceedings of WebNet 96—
`World Conference of the Web Society
`San Francisco, California, USA; October 15-19, 1996
`
`AACE ASSOCIATION FOR THE ADVANCEMENT OF COMPUTING IN EDUCATION
`
`Google Inc.
`GOOG 1008
`CBM of U.S. Patent No. 6,286,045
`
`Page 1 of 11
`
`
`
`Copyright © 1996 by the Association for the Advancement of Computing in Education (AACE)
`
`All rights reserved. No part of this book may be reproduced, stored in a retrieval system, or
`transmitted, in any form or by any means, without the prior written permission of the publisher.
`
`The publisher is not responsibile for the use which might be made of the information contained in
`this book.
`
`Published by
`
`Association for the Advancement of Computing in Education (AACE)
`P.O. Box 2966
`Charlottesville, VA 22902 USA
`
`Printed in the USA
`
`Page 2 of 11
`
`
`
`FULL PAPERS
`
`BRIAN BEVIRT
`Designing hypertext navigation tools
`MARTIN BICHLER & STEFAN NUSSER
`Developing structured WWW-sites with W3DT
`JOHN BIGELOW
`Developing an Internet section of a management course: Transporting learning Premises Across Media
`MICHAEL BJORN & YIN YUE CHEN
`The world-wide market: Living with the realities of censorship on the Internet
`MANFRED BOGEN, MICHAEL LENZ & SUSANNE ZIER
`A Broadcasting Company goes Internet
`WILLIAM A. BOGLEY, JON DORBOLO, ROBBERT O.ROBSON & JOHN A. SECHREST
`New pedagogies and tools for Web-based calculus
`ELLEN BORKOWSKI, DAVID HENRY, LIDA LARSEN & DEBORAH MATEIK
`Supporting teaching and learning via the Web: Transforming hardcopy linear mindsets into Web flexible
`creative thinking
`ALEXANDRA BOST
`The WWW as a primary source of customer support
`WOLFGANG BROLL
`VRML and the Web: A fundament for multi-user virtual environments on the Internet
`CHRIS BROWN & STEVE BENFORD
`Tracking WWW Users: Experience from the Design of HyperVis
`PETER BRUSILOVSKY, ELMAR SCHWARZ & GERHARD WEBER
`A tool for developing adaptive electronic textbooks on WWW
`ROBERT BUCCIGROSSI, ALBERT CROWLEY & DANIEL TURNER
`A comprehensive system to develop secure Web accessible databases
`ANTONIO CAPANI & GABRIEL DE DOMINICIS
`Web Algebra
`SCOTT CAPDEVIELLE (BUSINESS/CORPORATE)
`Capturing the state of the web
`CURTIS A. CARVER & CLARK RAY
`Automating hypermedia course creation and maintenance
`LEE LI-JEN CHEN & BRIAN R. GAINES
`Methodological issues in studying and supporting awareness on the World Wide Web
`CAROL A. CHRISTIAN
`Innovative resources for educational and public information: Electronic services, data and information from
`NASA’s Hubble Space Telescope and other NASA missions
`BETTY COLLIS, TOINE ANDERNACH & NICO VAN DIEPEN
`The Web as process tool and product environment for group-based project work in higher education
`MARGARET CORBIT & MICHAEL HERZOG
`Explorations: The Cornell Theory Center’s online science book
`CHANTAL D’HALLUIN, STEPHANE RETHORE, BRUNO VANHILLE & CLAUDE VIEVILLE
`Designing a course on the Web: The point of view of a training institute
`P.M.D. DE BRA
`Teaching hypertext and hypermedia through the Web
`DIETER W. FELLNER & OLIVER JUCKNATH
`MRTspace: Multi-user 3D environments using VRML
`BARRY FENN & JENNY SHEARER
`Delivering the daily Us
`RICHARD H. FOWLER, WENDY A.L. FOWLER & JORGE L. WILLIAMS
`3D visualization of WWW semantic content for browsing and query formulation
`
`Page 3 of 11
`
`
`
`Tracking WWW Users: Experience from the Design of HyperVisVR
`
`Chris Brown
`Communications Research Group
`University of Nottingham, University Park
`Nottingham NG7 2RD ENGLAND
`ccb@cs.nott.ac.uk,
`
`Steve Benford
`Communications Research Group
`University of Nottingham, University Park
`Nottingham NG7 2RD ENGLAND
`sdb@cs.nott.ac.uk
`
`Abstract: We are generally concerned with the development of a hypertext database visualization
`framework, HyperVisVR, which supports3D collaborative visualization, dynamic responsiveness to
`database updates and representation of the underlying database users in the visualization. In developing
`a plug-in database access module for the World Wide Web (WWW), we have needed to solve several
`problems related to the WWW's loosely-coupled stateless architecture. One such problem is that of
`tracking WWW users as they move between pages and servers. This paper discusses current
`approaches to tracking WWW users, proposes new ones and explores issues of privacy and mutuality
`with respect to the monitoring of hypertext access.
`
`Introduction
`
`As the WWW increases in size and complexity, there becomes a need for advanced tools to manage
`information overload. Image maps, navigation bars, server push/pull, Java, ActiveX and Shockwave are among
`the layout techniques that are being used by content providers to generate the site designs which predominate
`the WWW today. Although these complex web site layouts may improve navigability within the site itself,
`when we consider the WWW as a global distributed hypertext database, they can cause confusion as browsing
`users are met with a bewildering array of non-standard navigational tools and controls. Despite this recent
`proliferation of these non-standard interfaces, hypertext navigability is not a new problem. In 1990, Jacob
`Nielsen addressed the issue of users becoming lost in hyperspace and proposed the use of overview diagrams
`(maps) and fish-eye techniques [Nielsen 1990]. These approaches have been adopted by several researchers
`trying to improve the navigability of the WWW, particularly Dmel with his WebMap tool [Domel 1994], and
`Mujherta with the Navigation View Builder [Mujherta 1994]. The motivation for this paper is a project entitled
`HyperVisVR, which is a framework for visualization of large hypermedia databases within a 3D virtual world.
`The driving goals of this framework are:
`
`• Extensible through plug-in modules to support new databases and visualization styles.
`• Fully dynamic visualization which is responsive to viewers movement through virtual world.
`• Database access is represented in the visualization: users moving through the database have
``embodiments'.
`• Modifications to the database are reflected in the visualization.
`• HyperVisVR applications support peer-peer communications to represent other visualization viewers in
`the visualization itself, and to share cached information about the database objects.
`
`As an underlying database for HyperVisVR, the WWW has the benefit of being ubiquitously available
`throughout the world. However, we believe it poses some unique problems due to its loosely-coupled, stateless,
`architecture. This makes it a particularly interesting first target for our visualization framework. To be more
`specific, the problems with the WWW as a database for HyperVisVR are:
`
`•
`
`It is difficult to extract rich meta information about objects on the WWW.
`
`Page 4 of 11
`
`
`
`• Objects and links are stored together, making it difficult to analyse the structure of the WWW without
`performing an exhaustive search through the objects themselves.
`• As a connectionless, stateless protocol, HTTP makes it very difficult to track users and their actions on the
`database on a global scale.
`
`However, we don't believe that these problems are limited to HyperVisVR and many researchers are trying to
`find ways around the limitations on HTTP as a hypertext database protocol, and HTML as a hypertext object
`description language. Finding backwardly-compatible solutions to these problems would be of benefit to the
`research community as a whole. In this paper, we turn our attention to the specific problem of tracking users of
`the WWW as they browse and search the global WWW infrastructure. Potential applications of our ideas
`include:
`
`• WWW providers might optimise their servers by analysing the paths of users through their servers.
`• WWW providers might provide their clients with more detailed usage logs.
`• WWW applications might provide a mutual awareness of users leading to opportunities for encounters.
`• Tools to help users understand the relation between their current location and its surrounding context.
`• Search tools might use the activity of users around a set of WWW resources to indicate interest in them.
`
`User Tracking
`
`We now focus on the problem of tracking users as they pass through the WWW. Users can be tracked at three
`different places. A server can track users moving through it, a browser can be modified to provide tracking
`information whenever a database access is made and proxy servers can track the users who use the proxy to
`access other areas of the WWW. The first two sub-sections below assume that all connections are direct
`between the user agent and the origin server. Proxies further complicate user tracking techniques, and are
`considered in section 2.3
`
`Tracking at the Server
`
`The key reason that user tracking at the server side is so difficult is that HTTP is stateless. Each request for an
`object from an HTTP server is completely separate, and cannot easily be associated with previous requests from
`the same user. The following sections discuss methods for user tracking that are available with current and
`proposed versions of HTTP, extensions to browsers that enable user tracking, and suggestions for
`improvements to HTTP which would make user tracking cleaner and more reliable.
`
`In HTTP/1.0
`
`There are several request headers specified in HTTP/1.0 [Berners-Lee et al. 1996] which can be used by the
`server to link up requests into a click-trail for a particular user in a given session. These are: From and
`Referer[sic]. In addition, the IP address of the machine generating the request can be used as a basis for
`tracking. The From request header is an optional field, specified to contain the Internet e-mail address of the
`user. However, only several less popular browsers actually make use of this field, as it is regarded by most to be
`a breach of the privacy of the user to send his personal information in each request. From fields are most
`widely used in WWW robots to identify the administrator in case there is a problem with the robot. The
`Referer request header can also help track click paths. This field can be used by the browser to specify the
`address of the resource from which the current request address was obtained. This can be interpreted by the
`server to link a previous request to the current one.
`
`As browsers follow outside links from the server to other servers, their movement can be tracked through the
`use of cgi-redirect scripts. All outside links on a server are modified so that they link to a local cgi-redirect
`script, and pass the location of the outside link to the script as a parameter. The cgi-redirect script can log the
`movement, and then issue a redirect to automatically route the browser to the outside resource.
`
`Some sites use intelligent algorithms which analyse log files and link up requests into click-trails through the
`server using a combination of IP address, the From and Referer request headers (if present), a maximum time
`
`Page 5 of 11
`
`
`
`between subsequent requests, and an understanding of server structure to recognise which items a user is likely
`to access based on his current position in the server. It may be possible to do this analysis in real-time as each
`request was served, but it would pose a large overhead to already over-loaded servers.
`
`Authentication is often used by sites needing a fail-safe and universal way to track users through their servers.
`The advantages of this method are that it provides a method of user tracking which works with most browsers
`and it identifies individual users on subsequent visits even if they are connecting from a different computer.
`However, it is extremely inconvenient as it requires users to first register with the site and then remember a
`username and password that they should use on all subsequent visits to the site. Many users find that it
`generally isn't possible to use the same username and password at all such sites as many won't allow users their
`primary choice of username/password. Furthermore, based on our own experiences, we expect that many users
`who would otherwise look through the site may be dissuaded because of the inconvenience of registering or
`remembering their username. Thus, in general, we would not recommend this technique for sites hoping to get
`a high volume of traffic and who wouldn't otherwise use authentication for controlling access to their site.
`
`HTTP/1.0 Extensions
`
`The Cookie mechanism for client-side stateful transactions in HTTP is an extension to the HTTP protocol
`proposed by Netscape Corporation and implemented by the Netscape browser and several servers [Cookies
`1995]. When a browser requests a resource from a server for the first time, the server responds with a cookie,
`which the browser stores and sends as part of each subsequent request. This allows the server to link up
`requests from a particular browser into a click-trail. Cookies can be persistent, linking requests from one
`browsing session with requests from the previous one.
`
`The Keep-Alive extension to HTTP/1.0 allows several resources to be requested over a single connection. This
`is implemented by the Netscape browser and several servers. In principle it allows several requests to be
`matched up together as coming from one browser. However, in practice browsers use it only for single pages,
`requesting the page itself and all its embedded objects in one request. This limits the usefulness of the
`extension to follow a browser between distinct pages.
`
`In HTTP/1.1
`
`The HTTP/1.1 proposal introduces a new persistent connection architecture as the default connection type
`[Fielding et al. 1996]. This supersedes the Keep-Alive extension header described in 2.1.2. Any number of
`requests can be made on a single connection, until either the server or browser closes the connection. The
`specification does not make clear the circumstances under which connections should be closed or maintained
`and as, at the time of writing, there are no widespread implementations of the protocol, it is difficult to
`comment on whether this new architecture will improve user tracking. It is likely, however, that for matching
`of click-trails where the requests are punctuated by hours or days, that this architecture will not help.
`
`Tracking at the Browser
`
`The chief difficulty with server-side click-trail tracking using any of the mechanisms described in section 2.1 is
`that you can only track requests to the server. Frequently, browsers cache pages, and provide history
`mechanisms to allow navigation `Back' to the previous and `Forward' to the next page in the history. The
`browsers quite correctly do not generate new requests for this navigation. A consequence of this is, however,
`that it is not possible to maintain an accurate position of a user within a site if the user has navigated using the
`history mechanism. This is a major problem as a study at Georgia Institute of Technology analysed browsing
`strategies and determined that a total of 42.7% of navigation was through the history mechanism [Catledge et
`al. 1995].
`
`An alternative to user-tracking at the server side is to extend users' browsers to send usage information to
`interested parties whenever a new page is accessed. This can be implemented very easily with Mosaic's
`Common Client Interface (CCI) and a small helper application which connects to the CCI port of the browser,
`and relays WWW movement information via TCP or multicast to interested parties. Both WebCast [Burns
`1995] and FollowWWW [Brown et al. 1996] are applications which make use of this technique. An alternative
`would be a Netscape plug-in which monitors the actions of the user and sends movement information.
`
`Page 6 of 11
`
`
`
`The major disadvantage of these schemes is that they need to be explicitly configured by the users. The
`Netscape plug-in needs to be downloaded and installed; the Mosaic CCI application needs to be downloaded,
`installed, and then Mosaic needs to be configured and the helper started at the start of each session. It is likely
`that due to this inconvenience few users will download and install the tracking software, unless some form of
`incentive is given (free access to a subscription service, etc.). Furthermore, there are then the problems shared
`by any public software developer: supporting multiple architectures, user support, fixing bugs and notifying
`users of updates.
`
`Commentary
`
`The loss of real-time server-side tracking accuracy due to history mechanisms described in Section 2.1 might
`suggest that a browser-oriented tracking mechanism would be more desirable. However, whilst browser-side
`tracking is certainly more accurate, it suffers from scalability problems, and potentially low uptake due to the
`explicit configuration required. It may be possible to increase the accuracy of server-side tracking with certain
`browsers through several techniques. One such technique is the use of an anchor to an unavailable background
`image, embedded in the page. Netscape and Mosaic's history mechanism causes all unloaded images to be re-
`requested whenever the page is revisited, so careful analysis of requests for this unavailable image, including
`the Referer header, will indicate when a particular page has been revisited through the history mechanism.
`
`A hybrid approach which combines server-side and browser-side tracking would be for the server to include a
`reference to a small Java tracking applet with each page. The applet would have the sole responsibility of
`contacting the server each time the user departs a page or navigated back to a page through the history
`mechanism. The applet would be stored in the browsers cache, and so wouldn't be loaded across the network
`for each page. The Java applets may even be able to track the browsing activity levels on the workstation to
`determine if a user is actively viewing a page. Doing so, however, might be regarded as a breach of privacy.
`
`Both of these approaches could be handled within the WWW server which issues the resources, or they could
`be served by `tracking servers', in a similar way to the current proliferation of `page counting servers'.
`Tracking servers could handle page tracking for a number of different sites. They could act standalone for their
`benefit only, or make the tracking information available to applications such as HyperVisVR through TCP,
`UDP or multicast communication.
`
`Cache/Firewall problems
`
`The need to accurately track the movement of browsers clashes horribly with the application of proxy cache
`servers and firewalls to drastically reduce the amount of network bandwidth consumed with redundant
`requests. The crux of the problem is based on the load-based algorithm proxy caches use to determine how
`long a particular object should be cached for. This results in the original WWW server `seeing' extremely few
`requests from proxy servers for its most popular resources (as the caches store these popular objects), and an
`disproportionately high volume of requests for its least popular resources. As the popularity of proxy caches
`increases, this could completely invalidate the use of visualization such as HyperVisVR, which relies heavily
`on usage and popularity information. Possible solutions to the proxy cache problem can be broadly categorised
`into `ignoring the cache', `beating the cache', and `working with the cache'. We now discuss each of these in
`turn.
`
`It is possible to completely ignore the WWW population which accesses the server from behind a proxy cache
`by disregarding all requests which contain a `Via' header or the word `via' in the User-Agent header from the
`cache. Ignoring these misleading requests would seem like a good, straightforward approach to the problem.
`However, it is not possible to assume that the `direct access population' will be a representative sample of the
`entire requests. Many WWW users access through a cache because of organisation rules or country-based
`bandwidth problems, so by eliminating these users from the statistics you could unwittingly be excluding whole
`classes of users from the tracking statistics.
`
`Many content providers have resorted to `beating the cache' when attempting to obtain full access statistics and
`tracking information. HTTP/1.0 specifies a `Pragma: no-cache' header, which is an instruction to the cache
`
`Page 7 of 11
`
`
`
`server not to cache that object. However, due to the misuse of the header by content providers, many proxy
`cache administrators have resorted to ignoring the header and caching the resource anyway. Cache busting
`methods include appending a random segment to the URL which confuses the cache into thinking that all the
`resources are different, and generating all pages dynamically through a cgi-script, the results of which do not
`get cached.
`
`Working with the cache implies that you allow the proxy to cache your resources, and you are supplied with a
`log of cache-served resources originating from your server. This is not feasible without some form of
`automated system, as the number of proxy-cache servers is huge, and the structure of the proxies is completely
`uncontrolled. Such systems are under development, and when ready, it may be in the proxy cache's best interest
`to supply such information, as it might help the cache-busters in their defection from enemy to ally! To send
`information about each request to the origin server of every resource in the cache would severely impact on
`Internet bandwidth, and reduce the benefit of running a cache in the first place. It may be feasible however to
`consider some form of periodic batched transfer of information to selected origin servers that requested the
`information by setting a special header when the resource was first requested from the server.
`
`It is worth taking a paragraph to consider the impact of proxy caches on the unavailable image and Java applet
`solutions presented above. Based on experiments with NCSA httpd 1.5.1, proxies do not cache 404 unavailable
`responses, which means that this solution will work to track users behind proxy caches and firewalls.
`Furthermore, it will not impact on the network as much as cache busting algorithms, because no data is being
`transferred - only the request for the unavailable image. Java applets are cached by the proxy, but they are still
`allowed to communicate directly with the tracking server. There have been some problems with running
`networked Java applets from within Netscape from behind firewalls. At the time of writing, the author knows
`no solution to this particular problem.
`
`This concludes our discussion of current and possible techniques for tracking users as they access the World
`Wide Web. The following section now briefly discusses some of the ethical issues that arise as a result of this
`idea.
`
`Privacy and Ethical Considerations
`
`Techniques for tracking users, as discussed in this paper, raise a number of ethical issues concerning privacy
`and security. Such issues are complex and hardly ever clear cut. However, they are also of great importance
`and so warrant discussion and consideration during the technical development process. We begin by
`identifying the kinds of information that might be gathered about the presence and activity of users on the
`WWW. These include:
`
`• Monitoring general access trends - recording patterns of access by groups of people.
`• Anonymous monitoring of individuals - recording details of an individual's access but ignoring their
`identity.
`• Non-anonymous monitoring of individuals - also recording the identities of people accessing the WWW.
`• Persistent recording vs instantaneous awareness - deciding whether monitoring information is recorded for
`subsequent storage, analysis and use or whether it is only made available at the time of access (e.g. to
`enable chance encounters and stimulate social interaction).
`
`There are may also be many possible uses of such information. For example:
`
`• Making colleagues generally aware of each other's presence in much the same way the shared buildings
`and offices support the coordination of activity through casual awareness between their occupants.
`• Encouraging chance encounters between people browsing the same or related information.
`• By information and service providers in order to enhance services (e.g. developing new paths through
`information based on analysis of patterns of use) or as part of billing and accounting.
`• To enhance security by providing better awareness of who is accessing which sites and information (in the
`same way that video surveillance improves the security or many urban areas).
`It may be made available to third parties such as advertisers and government agencies.
`
`•
`
`Page 8 of 11
`
`
`
`It is important to point out that this kind of tracking may have clear positive benefits in some circumstances,
`but may have a negative impact in others. For example, information about an individual's preferences and
`activities might be used both to their benefit (to tailor systems to their needs) and also to their disadvantage (to
`compile personal profiles). Consequently, we propose three principles to guide the application and use of
`tracking techniques:
`
`• Notification - visitors should be notified in advance of the mechanisms and policies that are in operation at
`a given site so that they can decide whether to visit or not.
`• Mutuality - in general, systems should be designed so as to provide some degree of symmetry or mutuality
`as to awareness of presence. Thus, visitors to a site who are being monitored should be aware of the
`presence of their observers. Furthermore, it might be sensible to maintain a rough balance between the
`levels of information known by each party.
`• Balance of power - both the observer and observed should be able to influence the level of awareness or
`monitoring. In particular, it might be argued that people should not be made visible without their consent
`and should not be able to become invisible without the consent of those around them.
`
`Related Work to HyperVisVR
`
`Related work fits into three main categories: visualizing hypertext databases, visualizing access to hypertext
`database servers, and mutual awareness mechanisms for users simultaneously accessing hypertext resources.
`Visualizing hypertext databases has been a hot topic at WWW and hypertext related conferences. The
`approaches can be grouped into two categories. The first is those that follow a user through the WWW and
`creating visual maps of browsing history, such as [Ayers et al. 1995], [Domel 1994] and [Takano et al. 1996].
`Slightly different from these is HyperSpace which uses a 3D map to display browsing history, however it does
`not update the visualization automatically to show the current browsing position [Wood et al. 1995]. Another
`area of hypertext visualization research concerns the collection of a dataset from the hypertext database to
`visualize directly. 2D visualisations include venn-diagrams [Ralha et al. 1995], graphs [Mukherjea et al. 1995]
`and a distance-representing relief structure [Giradin 1996]. Virtual terrains (2.5D) have been used in the
`Hyper-G Harmony browser [Andrews 1995]. Full 3D visualisations include WebViz and GopherVR. WebViz
`uses a batch oriented approach to retrieve documents from a WWW server, parse them into a Hyperbolic tree
`structure, and display them [Munzner et al. 1995]. GopherVR is a 3D spatial interface to the Gopher system
`[McCahill et al. 1995]. Other research has focused on visualizing the access patterns on a particular server.
`Lamm et al. discuss a system which displays a globe in virtual reality, and maps the level of accesses to their
`WWW server from a particular geographical region onto the height of a bar extending from the appropriate
`location on the virtual globe [Lamm et al. 1996]. However caches, firewalls, and service providers upset their
`technique, as they mask the true origin of the browser.
`
`Turning briefly to some of our own work in this area, the Internet Foyer is a virtual reality WWW visualization
`of a set of pages, with visual representations of browsing users moving over the structure as they access the
`pages and move between them. It supports mutual awareness of WWW users on the same or similar pages
`concurrently. It also has a link to the real world: a real-time image of the visualization is projected onto the
`wall of a real foyer, and a video wall within the visualization allows VR users to see back into the real foyer.
`As such it links three spaces: the real world, VR and the WWW, and can be though of as a "mixed reality"
`[Benford et al. 1996]. The Internet Foyer was the precursor of HyperVisVR, which adds fully dynamic
`visualization, peer-peer networking, generic hypertext database support and plugin visualization styles.
`
`References
`
`[Andrews 1995] Andrews, K. (1995). Visualizing Cyberspace: Information Visualization in the Harmony
`Internet Browser. First IEEE Symposium on Information Visualization, Atlanta, GA, Oct. 1995, pp. 97-104.
`
`[Ayers et al. 1995] Ayers, E.Z., & Stasko J.T. (1995). Using Graphic History in Browsing the World Wide
`Web. Fourth International WWW Conference, 1995.
`
`Page 9 of 11
`
`
`
`[Benford et al. 1996] Benford, S.D., Brown, C.C., Reynard, G.T., & Greenhalgh, C.M. (1996). Shared Spaces:
`Transportation, Artificiality and Spatiality. To appear in Proc. CSCW'96, Boston, November 1996, ACM
`Press.
`
`[Berners-Lee et al. 1995]: Berners-Lee, T., Fielding, R., & Frystyk H. (1995). Hypertext Transfer Protocol 1.0.
`This is a work in progress, expiring on August 19, 1995. <draft-ietf-http-v10-spec-05.txt> (this is now RFC
`1945)
`
`[Brown et al. 1996] Brown, C., Benford, S., & Snowdon, D. (1996). Collaborative Visualization of Large Scale
`Hypermedia Databases, in CSCW and the Web, Proceedings of the 5th ERCIM/W4G Workshop.
`Arbeitspapiere der GMD 984, GMD, Sankt Augustin, April 1996.
`http://orgwis.gmd.de/W4G/proceedings/visual.html
`
`[Burns 1995] Burns, E. (1995). Collaborative Document Sharing via the MBONE.
`http://www.ncsa.uiuc.edu/SDG/Software/XMosaic/CCI/webcast.html
`
`[Catledge et al. 1995] Catledge, L.D., Pitkow, J.E. (1995). Characterizing Browsing Strategies in the WWW.
`Third International WWW Conference, 1995.
`
`[Domel 1994] Domel P. (1994). Webmap - A Graphical Hypertext Navigation Tool. Second International
`WWW Conference, 1994.
`http://www.ncsa.uiuc.edu/SDG/IT94/Proceedings/Searching/doemel/www-fall94.html
`
`[Giradin 1996] Giradin L. (1996). Mapping the Virtual Geography for the World Wide Web. Poster
`Proceedings of the Fifth International WWW Conference, 1996.
`http://heiwww.unige.ch/girardin/cgv/report/
`
`[Fielding et al. 1996]: Fielding, R., Frystyk, H., Berners Lee, T. (1996). Hypertext Transfer Protocol 1.1. This
`is a work in progress, expiring on December 7, 1996. <draft-ietf-http-v11-spec-07.html>
`
`[Lamm et al. 1996] Lamm, S.E., Reed, D.A., Scullin, W.H. (1996). Real-time geographic visualization of
`World Wide Web traffic. Computer Networks and ISDN Systems, 28(1996), 1457-1468
`
`[McCahill et al. 1995] McCahill, M.P., Erickson, T. (1995). Design for a 3D Spatial User Interface for Internet
`Gopher. Proceedings of ED-MEDIA'95, pp 39-44, Graz, Austria, June 1995. AACE.
`
`[Netscape 1995]: Netscape Corporation. Persistent Client State HTTP Cookies Preliminary Specification.
`http://home.netscape.com/newsref/std/cookie_spec.html
`
`[Mukherjea et al. 1995] Mukherjea, S., & Foley, J. (1995). Visualizing the WWW with the Navigational View
`Builder. Third International WWW Conference, 1995.
`http://www.igd.fhg.de/www/www95/proceedings/papers/44/mukh/mukh.html
`
`[Munzner et al. 1995] Munzner, T., & Burchard, P. (1995). Visualizing the Structure of the World Wide Web
`in 3D Hyperbolic Space.
`http://www.geom.umn.edu/docs/research/webviz/
`
`[Nielsen 1990] Nielsen, J.. The Art of Navigating Hypertext, Commuications of the ACM, 33, 3, pp. 297-310,
`ACM Press.
`
`[Ralha et al. 1995] Ralha, C.G., & Cohn, A.G. (1995). Building maps of hyperspace. Internet Multimedia
`Information: WWW National Conference, 1995, Minho University, Portugal.
`http://agora.leeds.ac.uk/spacenet/ghedini.html
`
`[Spero 199?]: Spero, S. Progress on HTTP-NG.
`http://www.w3.org/pub/WWW/Protocols/HTTP-NG/http-ng-status.html
`
`Page 10 of 11
`
`
`
`[Takano et al. 1996] Takano, T., Kubo, N., Shimamura, H., Matsuura, H. (1990). A Directory Service
`Architecture for the World Wide Web. Poster Proceedings of the Fifth International WWW Conference, 1996.
`
`[Wood et al. 1995] Wood, A.M., Drew, N.S., Beale, R., Heldley, R.J. (1995). HyperSpace: Browsing with
`Visualization. Poster Proceedings of the Third International WWW Conference, Darmstadt, Germany.
`http://www.igd.fhg.de/www/www95/proceedings/posters/35/index.html
`
`Page 11 of 11