Each chunk could be implemented to have different ratios of inodes to data, different block sizes, or other differences in layout, and can be initialized to a particular layout when first allocated. For example, a large file could be stored in a chunk containing exactly one inode and the associated data. File system growth is simple--add another chunk--and file system shrinkage is much easier, since the inode that ``owns'' any allocated block is in the same chunk (rather than located potentially anywhere on disk).
