Personally, I agree with the author. Since code executes linearly (in a single-threaded app), it makes sense to write it that way, too. We recently wrote a large a complex parser, interpreter, and backing database for a new text processing language we're developing and it was done entirely without classes or OOPS-type constructs. As the author suggests, it's mostly a lot of switch/case (or in our case, lisp cond expressions) that test for and handle the various cases that the parser might encounter, much as a human would. We've found this design much easier to reason about and debug than a bunch of classes that depend on each other.
