I think the biggest advantage is that it makes the data somewhat independent from the app. This sounds bad when you design the app, but in practice it's a godse

I think the biggest advantage is that it makes the data somewhat independent from the app. This sounds bad when you design the app, but in practice it's a godsend. Probably because it allows you to refactor the code as you wish, as opposed to having the data tying you up forever. I remember reading about the way Microsoft did things in office: saving was just dumping the program memory in a file. It's obvious it was bad, but what's interesting is that the most important reason it was so horrible is that it forced you to carry the old code with you forever, just to be able to load and save. Having another abstraction layer in between would have made it much more manageable, even if it saving would still mean dumping the memory somewhere. Of course, in their particular case that may have been an advantage.