In my experience, the key tenet for good database architecture is to assume that someone else is going to write a program, without your input, that's going to shove data into your DB's tables, without the involvement of your application layer. If the data that ends up in the database is wrong in such cases, then your application will be left holding the bag when it sees that data.

Much better to just tell your DBMS the rules you want enforced, such that when that ne'er-do-well (which might be, say, future you) attempts to insert crazy data, it just doesn't work.