I enjoyed the article on process partitioning, however the article advocating assertions in place of tests is very immature. It needs at least some explanation of how you might implement the advice, and how asserts would actually provide equivalent protection. I was very surprised when it ended, it feels incomplete.
You may have misunderstood his point or maybe I am. The way I understood, it is not about assertions such as
i = min < max ? min : max
but rather about the architecture of the software. The software should come built-in with the features that by default prevent bad software designs.
His argument is not to abolish tests and use asserts, but rather to define a set of cases for which you would not need to test because the framework, by its nature, makes it unnecessary.
His example was SQL injection. (a) If I were to write my own MySql queries which directly talk with the DB, then I would need to write tests around those query to make sure it is not possible to execute SQL injection. (b) If I am using an ORM that comes with feature so I don't write the sql queries, it creates it on its own and also comes with the feature to handles possible use-cases which would normally lead to SQl injection, in such case, there is no need to write the test.
