O wbudowanych testerach biernych

Rozważa się modele architektury rozproszonego, reaktywnego systemu zdarzeń dyskretnych, do którego zostaje dołączony bądź którego częścią staje się tester bierny. Modele adekwatne dla testowania aktywnego okazują się niespójne w kontekście testowania biernego. Identyfikuje się tu naturę tej niespójności i proponuje się nowy, bardzo prosty model, dzięki któremu tester bierny może zostać potraktowany jak każdy inny komponent systemu, co pozwala na naturalne modelowanie także systemów, w które tester bierny został wbudowany (zagnieżdżony).

In this paper a controversial concept of passive testing [2] is discussed, focusing on representation of a passive tester within the architecture of a distributed, reactive Discrete Event System. It is argued that well known architectural models used for paradigmatic active testing (Fig. 1) become deficient when applied to structures with a passive tester (Fig. 2), which undermines the basis for the formal treatment of passive testing. It is shown that such direct re-use either leads to intrinsically inconsistent and unimplementable interpretations (Fig. 2b and e), or requires additional mechanisms (co-location of a tester, Fig. 2c; instrumentation of a tested entity, Fig. 2d) that are restrictive and inconsistent with the specific use patterns of passive testing. A very simple, minimal architectural model is proposed (Fig. 3), in which each system entity is equipped with a pair of input and output ports, and communication channels are members of the Cartesian product of port sets. This model does not allow the inconsistent elements of previous models to be directly expressed, while making explicit their metaphorical interpretation (Fig. 4). It can deal with both external and embedded passive testers (Fig. 5), as they are treated exactly as any other system entity.