Currently, search queries are all POST requests resulting in the search results' pages being un-bookmarkable. This makes all the knowledge in OpenBiblio instances to practically be in the deep-web.
Using GET requests should partially remedy this by making it possible for users to bookmark and pass on references to sub-sets of the collections, i.e search results. In a sense, this provides a HTTP API for an instance.
It is a generally accepted that queries that do not alter the state of a datastore should be RESTful [https://en.wikipedia.org/wiki/RESTful].
Pairing this with a hyperlinked, crawlable trail of accessible web pages that collectively reflect the structure and content of the collection managed by an OpenBiblio instance should greatly increase the value of this system.