No, as long as it works.
> 2) Harmful to cache algorithms?
Potentially, yes. fs/inode.c:find_inode, for example, uses the inode
number to detect whether the inode is already in memory. So if the
same file is opened twice, you have to assign it the same inode number
- otherwise it will count as two files.
> 3) Nonworking?
Whether it works or not would depend on the semantics of your file
system. Maybe you can live with the restrictions. For example, when
you export your file system via NFS, the client gets an inode number,
which is used sometimes to relocate the file on the volume.
Maybe if you use some strong hashing algorithm (md5), you'll solve
some of the problems. OTOH, you will introduce new ones.
Regards,
Martin