Prototype Notes about the BROWSE and SELECT COLLECTION screen

 The browse collection screen provides an overview of th ecollections
on a timeline, selected filters, and a textual list of the collections
which can be sorted by date, topic, formats etc. There is tight
coupling between the 3 parts of the display. When the filters are used
the timeline and the list shows only the collections who answer the
query.  When the cursor is placed on a collection the timeline bar,
the title of the collection and the attributes of the collections are
highlighted. The short descriptionof the collection appears at the
bottom of the screen.
 After using the filter user can further reduce the set of collections
by manual deselecting them.  The remaining set of selected collections
becomes the scope of the search when using the "Search selected
collections" feature (see search tab on the left side of the screen).
This set of selected collection would also appear as the scope in the refined seach screen (but that's not working in this prototype).

 We used the data given to us by LC about 5 real collections and added
several other "fake" collections for testing purposes.

Each bar represents a collection and indicates the scope of the
collection. The interval field can be panned and zoomed by
manipulating the double-box slider beneath the timeline. The Y axis
has no meaning.  We considered Size of the collection for the Y axis
but the team was reluctant to define a size for each collection (is
size the number of items, byte size, time to evaluate?).  Instead, the
intervals would simply be iterated in a non-overlapping manner down
the vertical axis of the interval-field. Of course a better "packing"
algorithm will need to be applied as the number of collection grows.
Color-coding schemes for the intervals were also considered, and
rejected. Most of the attributes for the collections can assume
multiple values. For example, the Format attribute can be any
combination of Text, Film, Sound, etc., so a single color code would
not be appropriate.

The Java prototype has three main components: the Collection Overview,
the Collection Filters, and the Collection List.  The use of Java
allowed the Collection Overview and Collection List to dynamically
change in response to collection filters, showing only those that
satisfy the filter constraints. Furthermore, the Collection Overview
itself acts as a filter. Panning or zooming on a particular interval
of time specifies a temporal constraint, potentially filtering out
more collections.  The three interface components are linked via an
active cursor. Passing the cursor over a collection interval or
collection name highlights the other, as well as the appropriate
attributes in the Collection Filters. The URL for the collection
homepage is displayed in the status bar.  Pressing the mouse button
causes a jump to the collection homepage.  Each collection has its
attributes defined in a text file which is read by the Java Applet at
the start of the search and browse session. The file defines the name
and dates of each collection, and any number of attributes.
Collections can be edited, added, or removed, and the interface will
reflect these changes at the start of the next session.  The
Collection List can be sorted on any of the collection attributes.
The use of Java allows interface objects to be created dynamically,
based on the contents of a text file that defines the collections.
Java facilitates the active linking between components, links to
homepages, and collection sorting. Most importantly, Java facilitates
dynamic queries by allowing each discrete user event, such as a mouse
moving a slider, to be intercepted and processed, triggering a re-draw
of the visualization of results.  This tight coupling of user events
to graphical re-drawing is fundamental to dynamic queries.