Difference between revisions of "Mylyn/UI Design"
(Usage problems should be in bugs or on the Mylyn UI Nits page)
|Line 21:||Line 21:|
Revision as of 21:11, 23 August 2007
This page will be used to collect materials related to the design of Mylar's Focused UI. Please consider adding a screenshot of how you use Mylar to help inform the UI decisions discussed on bug reports.
While we are always tempted to add hierarchies to the Task List and there is never a lack of requests to do so, this is something that we have to proceed with very carefully because hiearchies are much dramatically more difficult to interact with and use than monocline lists. In general the Mylyn UI design attempts to follow the simpler monocline grouping metaphor over the hierarchical metaphor.
While there is never a shortage of demand for additional preferences, especially from advanced users, each preference that gets added has a significant cost in terms of UI complexity. Preferneces also mask ways in which the tool should change by providing work arounds particular to very specific use cases. We should strive to evolve the UI in a way that meets the use cases proposed by requested preferences in a way that does not overly clutter or add modalities to the UI. Oftentimes this can mean improving on or changing existing functionality instead of adding a preference to accomodate a specific use case. Since figuring out how to do that can require implementation, preferences can be added more liberally to the Sandbox component for the purpose of experimentation.
See "The Question of Preferences" in the following entry for a good discussion: http://ometer.com/free-software-ui.html
Using links vs. buttons in UI:
Uses of FilteredTree
- Should we include it in the Search view? Should Platform include it for all searches? bug 200411.
- The Task List currently has a content provider that supports guaranteed visibility (e.g. the query containing an overdue task will always show). The task contenxt model captures guaranteed visibilty for landmarks by maintaining a map. Platform's FilteredTree has a different, and somewhat expensive, mechanism of computing guaranteed visibility for matches and the viewer/filter level. Can we improve on this disparity?