in reply to Re: Reading the manual and knowing if you are getting good
in thread Reading the manual and knowing if you are getting good
This leads me to suspect that software test development is also not best when it designed on a black-box.
Amen to that.
Years ago I was responsible for defining the test plan for an OS/2 replacement for a previos DOS app. I was given the IT department generated (black-box mandated) test plan for the original as a starting point. It consisted of a stack of green&white lined fanfold about 7 inches thick. The first 1 1/2" of that were the tests for checking the handling of the applications configuration file.
Of the 20-something lines in the file, 14 were fully qualified path names to various other files. For each of these there was a comprehensive set of tests that consisted of manually editing the line of the config file to introduce particular errors (non-existant drive letter; non-existant directory; space in a directory name; space in a filename; non-existant file; etc. etc.) and then running the program and ensuring that it detected the errors. This batch of tests (the first 1 1/2") took 2 people a week to run.
Looking inside the code, it was obvious that
Given that the waterfall development method required that these tests be re-run every time the program was modified--even if the change was a totally trivial spelling correction; or something major but in a completely unrelated part of the program--the cost of that "black box innocence" to the reality of the inner workings was huge over the 7 years and hundreds of versions that had been tested.