Monday, 1 August 2011

Developer Journal 118 - On Testing, SVN Detective Work, and Various Ideas

I came in today to find that the simulation had crashed because it had used up too much computer memory, which is a good sign even though it indicates I'm not that great a programmer. The good sign is that the population I started from scratch is persisting.

The simulation ran for 2 hours 16 minutes and 45 seconds simulation time which took 7 hours 11 minutes and 10 seconds real time. That means that I can continue with my experiment of starting a population with learning and a population without learning. If I can this and other basic experiments and a good chunk of my thesis done, I can return to implementing more humanoid robotic bodies.

I've been using a free SVN service from Unfuddle but I've had to access it via wireless internet access through a wireless USB on my desktop at uni. I've recommend to students I'm tutoring to use SVN so I needed to check whether they can use it through the wired internet if they used SSL.

I tried but it didn't work.

I thought my version of SSL was out of date. I don't even know how to check. Move one. Come back to this later.

I thought my version of SVN was out of date. Nope, my version is 1.6.6, same as Unfuddle.

I thought it had something to do with "auth cache". Nope.

I thought it had something to do with "non-interactive mode". Nope.

I thought it was because I had to provide the username and password at the same time as I executed the checkout command. Nope.

I thought as above but the username and password had to precede the URL. Nope.

I thought that it has something to do with SSH. I don't know how to check. Come back to this later.

I thought I'd double check whether my free plan at Unfuddle had SSL.

My free plan does not have SSL! Grr! No wonder! You may think I'm was being silly not to check this earlier but the instructions at Unfuddle only mention this briefly. The way they show you how to access SVN through SSL makes you think you have SSL and that it's not hard to quickly try without thinking.

I opened an account at Project Locker which has SSL for free plans.

After many tries I realised that the username is your email address.

It works! Through my wired internet access and through my university firewall!.

Some ideas that have popped into my head lately.

AI idea. A service that can act as companions and competitors for games. Games, connect to the service and the service powers the non-player characters.

Release the executable after each session of work. Decide later whether to release the source code.

Games I have to remake, XCOM and then Darklands.

Two years theory seems to be about the right amount of time to aim for in creative projects. One year is much too rushed. Three years begins to feel like it's too long, creativity needs restrictions, the market has changed too much, and resource use and risk grows.

A number of my favorite works seems to follow this theory and it seems to benefit the process of improving with each project. If you follow the people who work on projects you can often seem the dramatic improvement they make from project to project and it seems the important thing is to actually regularly finish projects rather than finishing big projects. An example is the guys who worked on the Battlefield series of games. Their first game before Battlefield was very rough but they've improved dramatically with every game.


No comments:

Post a Comment