"If your interface requires instructions, it needs to be redesigned." ---Dan Rubin
Bad things about their old catalogue:
- lack of relevance ranking
- lack of permanent URLs/RSS
- siloing of collections
- lack of object type-appropriate behaviour
- inability to respond to user requests and suggestions
Data sources they want to bring all together:
- library catalogue
- theses and dissertations
- Google Books
- library digitization projects
- departmental digitizaton projects
- faculty research output
- archival finding aids
- licensed journals and databases
"Solr is the anti-silo."
"The 30,000 steamboats problem." How do you make sense out of piles of stuff that are all of the same, and overwhelm all the other results?
They use Cucumber to define and test how their relevancy rankings should work, giving specific cases and the expected results. When her librarians send her bug reports about rankings they send them in Cucumber format!!
On home page, they assume that most relevant result is an exact match on the title. On a special music place that's not true, so they boost the importance of names. Music students have such special needs that they did special stuff for them.
Plugin structure allows for local customizations without forking.