The criteria for deciding on the number of inputs are ultimately the risk that a failure could represent.
This risk can be broken down into such issues as section 16.8 in Chapter 16 and the possibility that
features interact. This is a matter of engineering judgment.
7.7.11 Partition the Input Space into Operations
You can further limit the input space in 4 ways:
??? By relating it to the operations of interest. The run types you have defined should share the input
space, to speed testing. Essentially there is a trade-off between a big operational profile and an
efficient test run. ???Efficient??? in this context means maximizing the possibility of finding bugs by
exercising more features rather than a greater range of variables.
??? By input variable ranges. We can, with some study, identify that some variables dominate within
a range or that they are non-uniformly distributed. Where we find such distributions we can
divide the runs by input space range.
??? By failure homogeneity. This means that within that range one failure (as determined by failure
behavior) is similar to another. Such failures will usually exercise the same code. The smaller the
run, the greater the homogeneity and the cost.
??? By risk. Irrespective of the probability that they be used, high-risk features must be thoroughly
tested. Lower risk levels can however be used to determine testing priorities.
7.7.12 Random Acts of Politically-Motivated Testing
The following should be randomized:
??? Data (within the required range), to avoid bias.
Pages:
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287