Technical <strong>Evaluation</strong> <strong>DICE</strong> <strong>Final</strong> <strong>Report</strong> The previous section highlights issues of concern to users, as well as indicating additional desirable features. There are other issues, mostly of a technical nature which should also be noted 7 , especially as these might impact on deployment (uptake), adoption <strong>and</strong> sustainability of the toolkit. (So, this section is written for technicians <strong>and</strong> other practitioners, as much as anyone). • Web Site The D+ website (http://devil.lib.ed.ac.uk:8080/dplus/) provides lots of useful information - links to relevant protocols <strong>and</strong> st<strong>and</strong>ards, use scenarios (case studies), background explaining web services, developer resources - but there are two (related) points which could be addressed to further help developers. First, though the site (context <strong>and</strong> framework pages) explains that the toolkit is based on OCLC <strong>and</strong> JAFER software, <strong>and</strong> the protocols used in target repositories, there is some confusion about claims relating to web services <strong>and</strong> federated searching on the home page. It is true that D plus offers federated searching through the Xgrain target, <strong>and</strong> for at least one target (Interlibrary) uses web service protocols (WSDL, Axis/SOAP), but this might be made more explicit. Second, as configured in the demonstrator, the toolkit largely simulates the SRU (<strong>and</strong> somewhat, SRW) protocols, which is not well explained in terms of the façade structural design pattern <strong>and</strong> service broker architecture. Though SRU may be a web service protocol, it is not be typical of what many developers might consider a web service to be, <strong>and</strong> D+ mostly simulates use of the protocol anyway. Search queries are constructed as a URL (SRU) from parameters entered in the interface, <strong>and</strong> the URL is parsed to pass those parameters typically as a Z39.50 query to the nominated target. The returned results are inserted in an SRW (xml) wrapper <strong>and</strong> rendered using the D+ style sheet. This approach is perfectly valid <strong>and</strong> at least goes some way to helping achieve interoperability; our concern is that the way the toolkit is presented may be confusing especially to practitioners unfamiliar with the idea of a façade. • Documentation As implied above technical documentation of the project could be improved in a number of ways, as it is currently rather abstract (high level). It would be helpful to have some brief narrative better describing: how the servlet is invoked; how search protocols are used; how the jsp pages interact; how xml transformations are effected using different style sheets (e.g. searchRetrieveResponse.xsl v searchRetrieveResponseZ.xsl files); navigation between result sets; certain configuration issues (considered separately) including a description of support files. The java code itself looks to be well written excepting that here <strong>and</strong> in the jsp code there looks to be quite a lot of hardwired code e.g. URLs for proprietary targets <strong>and</strong> xml code for style sheets. (Other issues relating to the code are discussed below). Output from the Tomcat console includes confusing (SOAP) messages relating to protocols which are not used, inherited from the base code. It might also be worth commenting that, depending on the metadata set supported, though results are returned in an SRW xml wrapper, the body of the xml might consist of particularly arcane structures e.g. GRS1 metadata. • Configuration Though the installation instructions given in the Developers Guide allow for ready deployment of the toolkit, there are a number of significant configuration <strong>and</strong> support issues outst<strong>and</strong>ing, such that the toolkit has not been seen to its best advantage. 7 See also Installation <strong>and</strong> configuration notes at http://www.icbl.hw.ac.uk/~santiago/elf.html (Toolkits) 14
<strong>DICE</strong> <strong>Final</strong> <strong>Report</strong> 1) Several targets in the downloadable version (release 2) did not work (ADLIB, COLIS), including key targets (British Library, Amazon, Google) which users would have found attractive. 15