I've a background in computer science and/but I found this section hard to follow. The distinctions between relaxed, restrictive and restraining mix concepts together which are really separate, namely
- the form of the rules and
- the way the rules are used, and thus the qualities of the grammar they generate.
Regular grammars have the tightest restrictions on the form of their rules but the grammars they generate impose the fewest restrictions on the use of a rule (together with context-free grammars). The idea that this situation is counter-intuitive results from this confusion of levels rather than from anything intrinsically puzzling about the kinds of grammar. Though I do appreciate that the author wants to highlight this distinction somehow.
I found it surprising to find the following statement: "Considering that it could be argued that Wikicode it is indeed a regular language." Whether Wikicode is or isn't a regular language seems to be more a matter of fact than opinion, but the statement suggests that there is some flexibility in the idea of a regular grammar. Is there a limit on how deeply nested constructs can be? If not, then Wikicode cannot be a regular language. Even if it were regular, the grammar needed to describe it might be far too difficult to understand and/or to implement to be of practical use or interest.
I hope these comments are constructive. I must confess to being a bit shocked that parsers were released for Wikicode before the issue of specifying a grammar for it was properly addressed, but then I am/was a theoretician first and foremost. Fairflow 10:17, 10 October 2011 (UTC)