Pointer cancellation

Success criterion 2.5.2, Pointer Cancellation, is Level A and has been in the standard since WCAG 2.1. For anything operated by a single pointer, one of four things has to be true: the down event executes no part of the function, or the function completes on the up event and can be aborted or undone, or the up event reverses whatever the down event did, or down-event activation is essential. In practice almost every interface satisfies it the same way, by activating on release, which is exactly what a browser’s own click event does: it fires on the up event, on the element the press started on, and going anywhere else before letting go cancels it.

The reason this is a criterion and not a preference is that down-event activation deletes the one recovery an imprecise pointer has. A tremor, a head pointer, a mouth stick, a shaking bus, a screen read through a magnifier: all of them produce presses that land on the wrong control, and the reader knows it the instant it happens. Sliding off before release is the recovery, and it is available for free right up until somebody writes a pointerdown handler that acts. That is where the failures live. They are almost never in ordinary buttons and almost always in hand-rolled ones: a canvas tool, a custom menu that commits on press, a card that navigates from mousedown because it felt faster.

Faster is the honest argument for the other side, and the criterion answers it: if the down event has to act, then the up event must reverse it, or an undo must exist. A control that previews on press and commits on release passes cleanly, since the preview is not the function. So do gestures where the press starts something rather than doing it, such as a long press, a drag past its drag threshold, or a hold to confirm that can be released early. Essential is narrow and mostly means the down event is the point, as in a piano keyboard where a key has to sound the moment it is touched.

Two neighbours are worth keeping apart from it. Click-through is about a click being accepted by a surface that only just appeared, which is a timing problem rather than a press-versus-release one, though both are about a tap the reader did not really aim. A ghost click is the same tap arriving twice through two event systems. And this criterion travels with the rest of the pointer family added or sharpened in recent versions, target spacing and the dragging alternative: between them they ask that a control be big enough to hit, reachable without a drag, and forgiving of the press that missed.