The library itself has full featured IO interface, i.e. open, save, and save-as. Though it is meant to be tiny, it has exception mechanism and two approaches to read the file. After several trials, I decided to use .NET's native deflate algorithm for its efficiency and convenience, though I don't know why NET doesn't have a build-in CRC32 hash algorithm.
You can find all the code at Codeplex.
I will add more compression algorithms as I fully develop the library.