The actual fishing component of their Fishing Game is all for show. Even Terry's old Mulefarm simulated fishing (i.e. 2 minutes of casting, caught __x__, repeat x 20). The actual fact is, it isn't required. There are only two important stages:

1. Starting the game. Gaia Online returns a packet of available fish (i.e. [0, 9, 12, 48]). That would translate to 0 rare, 9 large, 12 medium, 48 small.
2. You send the final packet to the server with the fish in your bucket. It will error if: a) The fish you send weren't in the original list Gaia sent you (i.e. you try and send a rare without it being in the list of available fish), and b) You sent the packet in less than 120 seconds after that original packet (not sure if it's exactly 120 seconds, but it's close).

The amount of large fish decreases the longer you play and, contrary to popular belief amongst users, rares are actually random (though how random depends on what bait you're using). A grade = 1/10,000. Next best is 1/100,000 and it's impossible to catch a rare with F grade bade.

I've simplified this all a bit, because there's a lot of flags that detect whether or not your hacking and then log it, but provided we simulate it all properly, we could create a client that doesn't rely on the swf. Of course, because they use reCaptcha, it couldn't be 100% automatic. I think the best recaptcha OCRs I've seen only have a success rate of 10-30%. However, if you guys wanted trk on this project, I'd be happy to go along with it (the user would just have to input the captcha manually).

--

As for re-designing Towns, Riddle has a fair point. 99.99% of users would never look at the code. The only reason I can think for them re-writing from scratch is to either make the final product better for the end user, or to make future collaboration easier and hence more efficient. I've never really looked at Towns - I just thought it was a place for people to walk around and talk?