Client: University of East Anglia

Project synopsis: An online pedagogic resource containing works from two museum collections

Project timescale: Two years

Specific requirements: Autonomous site that grows and adapts over time

Technology used: XML, XSLT, Perl, Apache Cocoon 2.0, PostgreSQL, Linux

References: Artworld, JISC 5/99

Artworld is one of the earliest Luminas web applications that makes use of Cocoon in a live context. Although old, it tackles many of the issues common to further- and higher-education websites even today, including integration with legacy applications, exposing "dark data" and using a clean, semantic design.

Artworld aimed to take content from collections of artifacts at the Sainsbury Centre for Visual Arts and the Oriental Museum, and make them available via the web within an interactive, searchable, self-sustaining archive. Users would not only be able to view images, but also to write about them and to view what others have written about them. In this way, a community would grow amongst the users and researchers of the site, giving it a vibrancy and fresh content beyond the timescale of the funded project itself.

The content for the site was prepared by the UEA project staff using a custom-written Microsoft Access database. Slides of the resources were scanned using Adobe Photoshop, then opened in Access which captured available image metadata (format, size etc.) and prompted the user to enter further metadata about the resource (such as title, provenance, etc.). This process was duplicated in Durham for resources at the Oriental Museum. Once the data was captured, an export from Access provided a number of CSV files that could be uploaded to the web server.

Exported metadata was ingested into the PostgreSQL database underlying the site using a perl script. PostgreSQL was selected as the most full-featured Open Source database available at the time of the project. One of the critical aspects of the Artworld solution was that the software should be capable of being deployed by anyone looking to provide a similar site. This ruled out expensive license options such as Microsoft SQL Server or Microsoft IIS. Issues of maintainability and future-proof data storage also added to the value of an open rather than proprietary database solution.

The web-based front end of the site was built using the Apache Cocoon web application framework, specifically the Cocoon 2.0 branch, which was the most recent stable release when development began. Cocoon was selected as the development platform because it was Open Source, because it supported XML, provided excellent separation of concerns, offered control of the URL space and semantic uris through the sitemap, was capable of publishing to multiple different formats (HTML, XML, RSS), made use of many open standards, and had proven to be fast and stable in previous sites.

The Artworld web application made use of many of the prominent features of the 2.0 version of Cocoon, particularly actions and XSP (eXtensible Server Pages). All presentation was kept in CSS and XSLT, to allow UEA's designer to make extensive changes to the look of the application in parallel to development of the content and logic.

The site was hosted on a Debian GNU/Linux Intel server, which represented the most cost-effective and open hosting solution available. Debian was selected due to it's excellent package management facilities and a reputation for security. Tomcat 3 was selected as the servlet container to run Cocoon in, with Apache httpd 1.3 in front to ensure continuous web availability even if the site itself was unavailable due to maintenance or errors.