My statement was about software, not about what can be learned from failed projects. Sure experimentation has value, even if it only leads to the insight that t

My statement was about software, not about what can be learned from failed projects. Sure experimentation has value, even if it only leads to the insight that theory can rarely be applied to software in materialized form. And I think that's the gist of the post. Take regular expressions, one of the theories that actually have been applied successfully to practice - because they are easy to understand and, as a corollary, describe practical concepts. Now, where are state machines implemented in their as a materialization of their abstract form? I don't know, maybe not even in a rather generalist tool like grep. They are at least not useful for lexers (one canonical use case for RE), because the formal overhead gets in the way of converting literals and looking up identifiers. In Software you need to cut corners to be able to implement useful functionality and achieve good performance.