High energy physics experiment triggers and the trustworthiness of software

For all the time and frustration that high energy physicists expend interacting with computers, it is surprising that more attention is not paid to the critical role computers play in the science. With large, expensive colliding beam experiments now dependent on complex programs working at startup, questions of reliability -- the trustworthiness of software -- need to be addressed. This issue is most acute in triggers, used to select data to record -- and data to discard -- in the real time environment of an experiment. High level triggers are built on codes that now exceed 2 million source lines -- and for the first time experiments are truly dependent on them. This dependency will increase at the accelerators planned for the new millennium (SSC and LHC), where cost and other pressures will reduce tolerance for first run problems, and the high luminosities will make this on-line data selection essential. A sense of this incipient crisis motivated the unusual juxtaposition to topics in these lectures. 37 refs., 1 fig