The usability of the CASCADE interface was assessed using a group of twenty students enrolled in a graduate level course in Interactive Systems. Logs of CASCADE usage were generated and the results were grouped into the following categories: fatal flaws, confusing conventions, limited support, or positive feedback. In the category of fatal flaws were a group of system crashes that were found to be caused by: users' double-clicking on buttons, by references to undefined environment variables, or by nesting comments to a depth greater than six.
Double-clicking is clearly related to the PC backgrounds of most of our novice users. The software was adjusted to account for this action. To overcome failures due to missing environment variables, logic was implemented to check on the existence of the variable. The failure due to nested comments involved the automated way in which files names are generated. Since names are concatenated and extended to include the new author's name and the new current time, the size of the string eventually exceeded the preset limit. This issue was addressed by allocating space on an as-needed basis.
Most of the confusing conventions were related to the novice status of the users with respect to workstation conventions, especially the X Windows System. These included the requirement to position the sprite within a window before entering text. The most commonly identified feature that generated confusion was the use of the right mouse button for comment link deletion. Alternative designs are currently being evaluated, including invoking a popup menu upon left mouse buttoning which would provide a menu of options incorporating the functions of all three current mouse buttons -- create comment, review comment, delete comment link. The naming of menus and submenu items generated some confusion among users. While this problem is minimized for frequent users, a review of menu naming conventions is underway. Similarly color conventions were confusing since no color legend was provided for interpretation. Again, as a system intended for regular users, we believe that the learning curve required will be more than compensated for by the vastly increased speed of preattentive processing.
The third group of short-comings was related to features that users would have
liked to have seen but didn't. Most if not all of these features will
ultimately be installed, but are not included yet, given the initial
development focus. These included: an UNDO feature,
context-sensitive help, a ``back'' button to be able to navigate to visited documents
in reverse order.