Slightly more pleasant, but not entirely unscary, I decided this week to actually finish the ad serving software I half-wrote back in 2002. Basically the scenario was that I needed to write something quick to serve a campaign I sold at the time. I wrote the serving part, but didn't write a single line of UI code, so I had to do everything by manually plugging stuff into the database.
The code wasn't horrible, but there was a lot of room for improvement and a better approach. I was able to refactor it pretty quickly into something I was comfortable with, and some initial testing shows it'll perform a little better.
I have a new found appreciation for reporting the statistics though. Millions of impressions create a lot of data. Data warehousing is not something I have a lot of experience with, but creating aggregate data on a regular interval seems pretty straight forward. The biggest decision is figuring out how much detail is really necessary. I'm guessing hourly stats are fine.
I know I've mentioned this before, but I wonder if revisiting old code ever becomes less scary.