I am involved in an effort to write a set of regression test scripts for a point of sale system. We are using ISO standards such as ISO/IEC/IEEE 29119-4 to determine our test techniques. For various operations I have been applying boundary value analysis.
For one specific operation on the till we can print out type balances (sales, refunds etc). The lower boundary is well defined as 0.00. However, the upper boundary is some magic number in the billions.
I could include a test script to say "test for this stupidly high number". However, I know that (through manual testing) it will be a huge effort. Many of these sub totals are limited in that the transactions have limits in themselves. For example, the system will only allow refunds of a maximum of 500.00 per transaction.
We can safely assume (by looking at client figures) that something like 99,999.99 would be a more realistic upper boundary.
Is it acceptable to re-define boundaries from functional to testable?
Or should I document the upper boundary as undefined/untestable, and capture a test scenario in some other test analysis approach (such as creating an equivalence class partition of "testable range" or use an "error guessing" approach to make sure such a large number doesn't break the system)?