Gert Schmeltz Pedersen is working on a search system called the Generic Search Service (or fedoragsearch) that has been endorsed by the Fedora team. The plan is to release this system with Fedora 2.2 (date unknown). This system abstracts over the actual search engine, and provides implementations based on both Lucene and Zebra.
The current design document indicates that the internal structure of this searching system will be somewhat like XTF, with configurable XSL transforms to prepare data for indexing and for display. However, this configuration will be done at the level of each indexing/display system, and not on a collection-specific basis. Note that the current design does not include any form of query processing. Queries are simply sent to the search engine as they are received.
A prototype of gSearch is available from the design document.
A sample searching interface provides an indication of the way URLs will be used.
We will want to index different keywords for each collection, so we will need a better way to manage the indexing process, and use a different stylesheet for each collection.
Current implementation notes
- Install gSearch in Fedora's Tomcat
- Make sure the config files in WEB-INF/classes are correct:
- absolute references for includes in config/rest/demo*
- Copy our foxmlToLucene.xslt to config/index/Lucene
- Use the Tomcat manager to reload the gSearch servlet
- See the instructions for Making a Collection Searchable