From Corrib Clan Wiki

Servlet-JSP Architecture for Search and Browsing in JeromeDL

1. I have identified 4 types of actions in JeromeDL (see Fig 1):
   1. preview resources
      • currently handled by preview.jsp
      • takes as optional parameters (type, creator)
      • presents list of resources in HTML version
   2. show concrete resource
      • currently handled by show.jsp
      • takes as an obligatory parameter (ID) of a resource
      • presents resource in HTML version
   3. search (simple, advanced, typed, semantic)
      • currently handled by search.jsp
      • takes various parameters
      • presents results as a list in HTML (special exception for semantic search)
   4. [MFB] multifaceted browsing (in production)
      • currently not supported - but should be included in the HTML presentation as additional action that user can perform on current set of resources
      • takes as an input list of filtering constrains build with facets and values from appropriate taxonomies
      • presents modified list of previous list of resources in HTML version

1. There a couple of requirements that makes the servlet-JSP architecture "a must"
   1. support for MFB - we need some abstraction layer to do that
   2. presenting resource/list of resources in other formats than just HTML (e.g. Marc Ont, DC, Bib Te X, RSS, ...)
      • this is currently done by applying MMS to single call for each resource
      • we need to support for the whole list of results in preview and search
      • the type of answer should be identified with parameter render=[html|marcont|dc|bibtex|...] appended to the URL of the call
   3. results of search and preview operation need to be divided into separated pages (e.g. 10 per page)
      • we could wait for the support from Sesame to do that
      • but we have to do it ASAP
      • and stay independent from the RDF Storage capabilities
      • additionally search features are build in JeromeDL so it would be hard to use sesame feature for paginating
   4. search results in JeromeDL might change over the time (semantic query expansion) what could be dangerous for presenting results in different formats and supporting MFB

1. the flow of actions in JeromeDL would look like that:
   1. user action (preview, show, search)
   2. processing the action to get:
      • RDF Query, or
      • list of resources (from search)
   3. Sesame produce results in RDF triples form
   4. RDF Triples are cached in temporary repository (memory or disk based)
   5. if user asked for HTML version (default):
      1. resources are sorted according to creator, title, ... (@Bugi can you do that in XSLT or you need to do that earlier?)
      2. resources are rendered to HTML with XSTL conversion
   6. if user asked for non-HTML view - MMS is used to deliver proper conversion
   7. if user decides to use MFB then his/her actions are translated into RDF Query but this query is performed either on cache or on whole repository (if he will choose relax query option). Please note that MFB can only be used in the HTML mode.

Fig 1: Servlet-JSP Architecture in JeromeDL Svg version