Interesting. One thing they leave implicit is the amount of work that parsing requires.

The result is that you can parse in time O(log n) if you use O(n) processors, so parsing requires O(nlogn) work. (This is probably to be expected.) I think this means that, if you have o(n) strings to parse, you're better off with the good old DFA solution which uses O(n) work and parsing the different strings in parallel.