Greetings and Salutations,

I'm currently on a grand quest to reinvent a wheel. Well, not really a wheel, a database. This quest exists for the purpose of discovering how to effectively structure data for simplicity of programming while attempting to maintain somewhat reasonable performance. I'm leaning towards a relational, table-based model because that's all I really have an understanding of at this point :).

My humble questions are as follows:

1. How would one persistantly store the data? All I'm doing right now is using a hierarchy of directories that basically goes rootdb/db/table.file. The table files consist of null-terminated rows. Is there a better way of structuring this?
2. I currently have no idea how to store meta-data such as column types and sizes and how to make use of them. I'm planning on writing this for only Linux & BSD systems if that makes a significant difference. Any ideas?

I've taken a moderately in-depth look around and found very little on the subject. All the books I've found seem to focus on data normalization and sql. Any advice or pointers to any Books or FMs that I can Read would be most excellent. Many thanks for your time and assistance (: