The Web Watch Project About Web Watch The
The Web. Watch Project About Web. Watch • The Web. Watch project is funded by BLRIC (British Library Research and Innovation Centre) • The Web. Watch project involves the development and use of web robot software for monitoring use of web technologies • Papers, reports, articles and presentations of the findings are produced by the Web. Watch project A Web. Watch Trawl A simple model of how the Web. Watch robot trawls communities is shown below Input file of URLs Resource A 100101101011 Web. Watch robot reads input file and retrieves 001010101 resources 10101101011 Summary file Resource A, B, etc. could be individual pages or entire websites 100101101011 Analysis and statistical programs produce 001010101 reports 10101101011 Resource B Report for UK Universities UKOLN is funded by the British Library Research and Innovation Centre, the Joint Information Systems Committee of the Higher Education Funding Councils, as well as by project funding from the JISC’s Electronic Libraries Programme and the European Union. UKOLN also receives support from the University of Bath where it is based.
Web. Watch Trawl of UK University Entry Pages Background The Web. Watch project carried out a trawl of UK University entry points on 24 October 1997. The trawl was repeated in 31 July 1998. Web Servers The most popular web server was Apache. This has grown in popularity, with a decline in the CERN, NCSA and other smaller servers. Microsoft's IIS server has also grown in popularity, perhaps indicating growth in use of Windows NT. Size of Entry Points The file size of HTML resource(s) (including frame sets) and images (but excluding background images) were analysed. Four pages were less than 5 Kb. The largest page was 193 Kb. The largest pages contained animated GIF images.
Web. Watch Trawl of UK University Entry Pages Web Technologies An analysis of some of the technologies used in UK University entry points is given below. Java and Java. Script Liverpool University is probably the only university entry page using Java None of the institutions trawled made use of Java. Subsequently it was found that one institution used Java. This institution used the Robot Exclusion Protocol to stop robots from trawling the site. Java provides this scrolling news facility Java. Script In October 1997 22 institutions used client-side scripting, such as Java. Script. By July 1998 38 institutions were using Java. Script. The University of Northumbria at Newcastle is one of about 38 institutions which use Java. Script is used to display picture fragments when the cursor moves over a menu option.
Web. Watch Trawl of UK University Entry Pages Metadata <META NAME="description" CONTENT="Mailbase is a national mailing list centre for UK HE"> <META NAME="keywords" CONTENT="mail", "listserve"> In October 1997 54 institutions used "Alta Vista" type metadata on their main entry point. By July 1998 the metadata was used on 74 entry points. In contrast Dublin Core metadata was used on only 2 pages on both occasions. <META NAME="DC. Title" CONTENT="The Mailbase Home Page"> <META NAME="DC. Creator" CONTENT="John Smith"> Possible Use of Alta Vista and Dublin Core Metadata Cachability Interest in cache-friendly web resources has grown since the introduction of network charging on 1 August 1998. Over 50% of institutional HTML resources were found to be cachable, with only 1% not cachable. Further analyses is needed for the other resources. A Web. Watch service is being developed to provide a web-interface to the telnet command, to give more helpful information. % telnet www. ukoln. ac. uk: 80 GET / HTTP/1. 0 HTTP/1. 1 200 OK Date: Fri, 28 Aug 1998 16: 22: 51 GMT Server: Apache/1. 2 b 8 Content-Type: text/html Telnet can be used to analyse HTTP headers, including caching information URL: http: //www. ukoln. ac. uk/ This resource uses HTTP/1. 1. The resource is cachable. The resource was last updated on … Possible Interface
Web. Watch Trawl of UK University Entry Pages Frames In July 1998 the following 19 sites used UMIST is an example of frames, compared with 12 in October 1997: a framed website • Essex • Bretton Coll. • UCE • Royal College of Music • Keele • King Alfred's Coll. • Middlesex • Nottingham Trent • Portsmouth • Ravensbourne Coll. • Teeside • Birkbeck Coll. • UMIST • Uni. Coll. Of St Martin • Thames Valley • Queen Margaret Coll. • Westhill • Scottish Agricultural Coll. • Kent Institute of Art and Design Liverpool University also uses frames but this was not detected by the robot due to their use of the Robot Exclusion Protocol. "Splash Screens" In July 1998 5 sites used clientside requests to provide redirects or "splash screens". De Montfort University displays a screen with a yellow background. After 8 seconds a new screen is displayed. "Splash screens" are created by <META HTTP-EQUIV="refresh" CONTENT="n; URL=xxx. html">
Web. Watch Trawl of UK University Entry Pages Hyperlinking Issues The Web. Watch trawls revealed some interesting hyperlinking issues, which are described below. Numbers of Hyperlinks The histogram of the numbers of hyperlinks from institutional entry points shows an approximately normal distribution. Six sites were found to have fewer than 5 links. One site contained over 75 links. Limitations of Survey The analyses do not give a completely accurate view for a variety of reasons: Discussion Many Links: • Provide useful "short cuts" • The address of one of the sites for experienced users with a small number of links was • Can minimise numbers of incorrectly given in the input file levels to navigate list (obtained from HESA). Few Links: • The analysis did not exclude • Can be confusing for new duplicate links. user • Can cause accessibility • Sites containing "splash screens" problems (e. g. for the were reported as having small visually impaired) number of links, although arguably the links on the second What is your view? screen should also be included.
Trends in UK University Entry Points Trawls of UK University Entry Points The Web. Watch project has surveyed UK University web site entry points on three occasions: 24 October 1997, 31 July 1998 and 25 November 1998. A summary of significant trends is given below. Server Usage The Apache and Microsoft web servers are both growing in popularity, at the expense of the CERN and Netscape servers, and a number of more specialist servers. "Splash Screens" The number of entry points using "splash" screen has increased from 5 (Oct 97), to 7 (Jul 98) to 10 (Nov 98). Metadata Usage Use of Dublin Core (DC) metadata grew during the summer 1998 from 2 sites to 11. DC metadata is still dwarfed by "Alta Vista" style metadata. Growth (Kb) Size Of Entry Points Trends in the sizes (HTML plus embedded images) have been analysed. The majority of entry points have not changed in size significantly, although one or two have grown (~ 100 Kb) or decreased in size (~50 Kb) substantially.
Web. Watch Services Web. Watch provides access to various tools and utilities which have been developed to support its work. These services can be accessed using a Web browser at the address <URL: http: //www. ukoln. ac. uk/web-focus/ webwatch/services/ >. HTTP-info Service A web form is available which can be used to obtain the HTTP headers sent when the resource is accessed. This service can be useful for getting information, such as the name of the server software, HTTP version information, etc. Doc-info Service A web form is available which can be used to obtain information on web resources. The Doc-info service is integrated with the HTTP-info service, enabled the HTTP headers are all objects contained in a resource to be analysed.
Web. Watch Technologies The Web. Watch project has made use of the following technologies: • • The Harvest indexing and analysis suite Perl for developing the Web. Watch robot Locally-developed indexing and analysis software A series of Unix Perl utilities for analysis and filtering the data • Excel, Minitab and SPSS for statistical analysis Trawling Software The Harvest software was used originally. Harvest is widely used within the research community for indexing resources. For example the ACDC project uses Harvest to provide a distributed index of UK. AC web resources. Unfortunately as Harvest was designed for indexing, it is limited in its ability to audit and monitor web technologies. The current version of the Web. Watch robot uses Perl. ACDC uses Harvest. See <URL: http: //acdc. hensa. ac. uk/>
Restricting Access Why Restrict Access? Administrators may wish to restrict access by automated robot software to web resources for a variety of reasons: • To prevent resources from being indexed • To minimise load on the web server • To minimise network load Robot Exclusion Protocol User-agent: * The Robot Exclusion Protocol Disallow: /images/ is a set of rules which robot Disallow: /cgi-bin/ software should obey. Typical robots. txt File A robots. txt file located in the root of the web server Issues can contain information on: Some issues to be aware of: • Areas which robots should not access • Particular robots which are not allowed access Web. Watch Hosts A robots. txt Checker Service • Prohibiting robots will mean that web resources will not be found on search engines such as Alta Vista • Restricting access to the main search engine robots may mean that valuable new services cannot access the resources • The existence of a small robots. txt file can have performance benefits • It may be desirable to restrict access to certain areas, such as cgi-bin and images directories.
Web. Watch Recommendations The final Web. Watch report makes a number of recommendations, based on its trawls, including advice for Information Providers, Web Administrators and Robot Software Developers Information Providers System Administrators Directory Structure The robots. txt File Directory structures can provide a form of metadata about a resource. It is recommended the information providers make consistent use of directories. Metadata The use of "Alta Vista" type metadata is recommended for use on key entry points. Frames can prevent indexing robots from accessing resources. If frames are used, there should be an alternative route to resources for robots. Web system administrators should ensure that web servers contain a robots. txt file. This may be used to restrict access to robots. HTTP/1. 1 Web system administrators should ensure that their server software supports HTTP/1. 1. Analysis of Robot Usage Web system administrators should periodically check log files for access by robot software. Software Developers Further Information Further recommendations are included in the final Web. Watch report. The report is available at <URL: http: //www. ukoln. ac. uk/ web-focus/webwatch/ reports/final/ >. Memory Leaks Memory leaks can cause problems, especially when accessing large nos. of resources. Robot software should include checkpoints, to facilitate restarts. User-Agent Negotiation Robot developers should be aware of server use of "User-Agent Negotiation" which may provide different information to robots and browsers.
Finding Out More About Web. Watch Ariadne Occasional Web. Watch reports are published in the online version of the Ariadne magazine. See: <URL: http: //www. ariadne. ac. uk/ issue 12/web-focus/ > <URL: http: //www. ariadne. ac. uk/ issue 15/robots/ > Publications The following Web. Watch articles have been published: • "Robot Seeks Public Library Web Sites" in LA Record, Dec 1997 Vol 99 (12) Web. Watch Staff • "Academic and Public Library Web Sites" in Library Technology, Aug 1998 The Web. Watch Officer is Ian Peacock (email • "Web. Watching Academic Library Web I. Peacock@ukoln. ac. uk). Sites" in Library Technology, Jun 1998 • "Web. Watching Public Library Web Site Ian's responsibilities include software development, running Entry Points" in Library Technology, the robot trawls, analysing the Apr 1998 data and producing reports. • "Public Library Domain Names" in The Web. Watch project is Library Technology, Feb 1998 managed by Brian Kelly • "How is My Web Community Doing? (email B. Kelly@ukoln. ac. uk). Monitoring Trends In Web Service Provision" in Journal Of Documentation, Vol. 55 No. 1 Jan 1999 The final Web. Watch report can be obtained from <URL: http: //www. ukoln. ac. uk/ web-focus/webwatch/reports/final/>
- Slides: 12