In stage 1 we use a glob pattern matching the file name. Stage 2 (which I will maybe implement some day) might use some additional magic byte sequence for file type detection.
There is a database named RavenDB. KrowLog starts with a K, which is a) distinctive and b) has an association to KDE.