Section: New Results

Transfer functions and latency

Participants : Géry Casiez, Alix Goguey, Stéphane Huot, Sylvain Malacria, Nicolas Roussel.

Our work on transfer functions mainly focused this year on edge-scrolling, which allows users to scroll a viewport while simultaneously dragging near or beyond its edge. Common implementations rely on rate control, mapping the distance between the pointer and the edge of the viewport to the scrolling velocity. While ubiquitous in operating systems, edge-scrolling had received little attention, though previous works suggested that rate control may be suboptimal for isotonic pointing devices (e.g. mice and touchpads) and space beyond the window's edge might be scarce, limiting scrolling control. To address these problems, we developed Push-Edge and Slide-Edge two position-based techniques that allow scrolling by “pushing” against the viewport edge [23] . A controlled experiment shows that our techniques reduce overshoots and offer performance improvements up to 13% over traditional edge-scrolling.

Figure 1. When selecting text with a touchpad, downward movements after crossing the viewport edge will (a) change the rate of automatic scrolling with existing techniques (rate control); or (b) manually scroll the document, stopping the pointer at the edge with push-edge scrolling (position control).

Our work on latency focused on its measurement in existing graphical user interfaces, a problem for which we developed a simple method [18] , [27] . Our method consists in positioning an unmodified optical mouse on the screen while displaying and translating a particular texture to fake mouse displacements, which results in controlled mouse events. This works with most optical mice and allows accurate and real-time latency measures up to 5 times per second. The method also allows easy insertion of probes at different places in the system to investigate the sources of latency. Measurements performed on different systems, toolkits and applications notably showed that latency is affected by the operating system and system load. Substantial differences were also found between C++/GLUT and C++/Qt or Java/Swing implementations, as well as between Web browsers.