February 28th, 2006

eyes black and white

Software Mistakes

It's deadline time at work, and I realize I do quite a lot of mistakes that usually involve jumping to conclusions. Jumping to conclusion is my biggest asset: I am able to find very fast the broken concept, even before I have all the information required to confirm the conclusion. However, running software doesn't require just correct concepts, it requires 100% detail-correct source code. Typos are incorrect. Modifications that you think you propagated to another CVS branch but didn't are incorrect (ouch for not using a better versioning system based on patch sets or history merges). But that's the way you learn. Do more testing, improve the tools to automate the testing. Too bad we don't have proper meta-level tools, and no time (and will) to build them. We should probably allocate explicit time for building such tools, when the deadline is further away...