I have 5.8.8, and I have 5.14.2. I've also taken the liberty of converting the string of a?'s into a{0,27}.

I recognize that you're trying to demonstrate pathologically bad behavior, but as written, you've got a pathologically bad regex.

Note the speed up from 24 & 30 seconds to seven one-thousandths of a second in both cases. Again, I like the idea of a regex engine with different design considerations than Perl's.

Your article would have been much better if it has used a class of pathological data with real world examples, and not used contrived examples that I rewrote in less time than it took to run.

So here's my results:

5.14.2:

     $ time perl -e 'print "aaaaaaaaaaaaaaaaaaaaaaaaaaa" =~ /a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?aaaaaaaaaaaaaaaaaaaaaaaaaaa/ '
     1
     real    0m24.227s
     user    0m24.210s
     sys     0m0.008s

     $ time perl -e 'print "aaaaaaaaaaaaaaaaaaaaaaaaaaa" =~ /a{0,27}aaaaaaaaaaaaaaaaaaaaaaaaaaa/ '
     1
     real    0m0.007s
     user    0m0.004s
     sys     0m0.004s

----------------------------------

5..8.8:

     $ time perl -e 'print "aaaaaaaaaaaaaaaaaaaaaaaaaaa" =~ /a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?a?aaaaaaaaaaa                               aaaaaaaaaaaaaaaa/ '
     1
     real    0m30.065s
     user    0m30.046s
     sys     0m0.016s

     $ time perl -e 'print "aaaaaaaaaaaaaaaaaaaaaaaaaaa" =~ /a{0,27}aaaaaaaaaaaaaaaaaaaaaaaaaaa/ '
     1
     real    0m0.007s
     user    0m0.004s
     sys     0m0.004s

--------------------------------