The DR-DOS/OpenDOS Enhancement Project

Only that 64-bit cluster addressing makes no sense with the FAT design. Even with the 28-bit addressing with 32-bit entries (FAT32) the FAT can hold up to 2^28 = 256 bin. mega entries. Such a large FAT would be sized 2^28 * 4 = 1 bin. GB. That's a lot, and I see no reason someone would want such a large FAT. Thanks to the FSINFO structure, cluster allocation and free space calculation would still mostly work in a reasonable time. Don't try defragmentation though.

An actual 32-bit FAT (i.e. 32-bit entries and cluster addressing) would contain up to 2^32 = 4 bin. giga entries and would then take up 2^32 * 4 = 16 bin. GB.

FAT24 with actual 3-byte entries would have been neat, up to 16 bin. mega entries taking up 48 bin. MB. (Of course FAT12-like cross-sector entries would show up again but I don't consider that a serious problem.)

It seems people are often calling exFAT FAT64 now - which is of course even more wrong and misleading than "FAT32" with actual 28-bit cluster addressing.