I can fold some of the native path stuff into cldc-full since a bunch of it will be handled via such means anyway. This makes things a bit simpler.
All I know is that for DOS, there is likely going to be many
InvalidPathExceptions thrown due to DOS limitations.
So for DOS paths there will just be more failures for most paths if code makes assumptions about filesystem limits. In other parts of the code, the constant pool initialization can be done as a simple linear sequence of stages of initialization sorted in processing order.