Malfador could have used an RDBMS for storing all of the data in the various AI files, component files, etc. and it would have probably worked more efficiently than parsing the text files, as is now done. The advantage to the text files is that anyone with a text editor can make their own modifications to the game. It would have been difficult (possible, but difficult) to provide the customization flexibility we have now using a database scheme to store all of the information.
If this were a production package where stability and data integrity were critical, I would definitely want it running on an RDBMS, but I believe that for a game of this nature, the text files are the way to go.
Having said that, I believe that the game would probably be more stable and perform faster if the data was in an RDBMS. What might have been cool, and might be good for SE5 would be to use the text files to populate an RDBMS. Then, people could make their changes to the text files and run a utility to load the changes into the database. Data validation would occur at that time, so you would never see an error crop up in the game because you hit an error in the text file. Multiple mods could easily be supported by adding a mod name to the key information for storing and extracting the data. I would imagine that this would entail a significant internal architecture change though, so we would not see anything like that in SEIV.
Ok, time to take off my designer's hat. It's too much fun to think about what could be done in a game when I am supposed to be debugging a GIS query interface that keeps wanting to lose the operator's selections.
