That's a very good question. Where should design decisions be documented? On one side, we should have requirements that we document with some high level TestCase. But whether the implementation is guided by time/memory efficiency, code simplicity, or just historically available packages is not easily described by tests. Obviously, you would have a hard time discovering this in legacy code (digging squeak wiki and mailing list?). The question should apply to Pharo too. Pharo is young, so maybe all design decisions can be retrieved by virtue of human memory... But will it be the case in 5 years?
We will introduce package manifesto and they should hold this. as well as false positive for rules checking and documentation. I think that any successful system will have such problem. Now high level rules, tests are documentation are important. Stef