ChromeVox has a few different ways of producing output. Sometimes, speech
is provided directly to Output.withString() -- after internationalizing
via Msgs.getMsg(). This at first seems like the right approach for most
situations, but the Output class has developed a sophisticated method
of automatically determining the output strings, based on the change in the
CursorRange that ChromeVox is focused on and the type of event that's
occurring (which could be navigation, or any
chrome.automation.EventType).
During navigation, output is generated from the current and previous ranges
without needing to be explicitly specified. It uses announcement patterns
that are specified in the file output_rules.js in a special markup language.
It also navigates the ancestors of both ranges to make appropriate announcements
for any nodes that have been entered or exited, again using rules from
output_rules.js.
When an AutomationEvent is received, output is produced by passing the event
and the current range. These announcements are also specified in
output_rules.js.