2) the 1024-bit "computer kilo" is only really unambiguous when combined with bytes. When applied to bits, some people mean SI-kilo and other 1024-kilo, which is a bit messy. At least it doesn't look as ugly as the KiBiBytes etc. that the lunix zealots want to impose on the world.
3) depends. An ASCII character is defined as one byte, where a byte is usually (but not always!) 8 bits. For a lot of languages, this is not enough, though, and then you have to use a
MBCS encoding (which can be unicode or "something else"). Unicode covers several different encodings, btw, and isn't just "two bytes per char".
5) sectors are imposed by the storage device, clusters by the filesystem. Since the storage device addresses in sectors, that's why
sectors go bad, and not clusters. A filesystem based on clusters will probably choose to mark the whole cluster bad, though. Not all FSes are cluster-based.
6) again, cluster size is defined by filesystems, not anything else.
7) when dealing with computers - and they aren't metric units (you probably mean
Système Internationale, btw?)

depends on the CPU, really. x86 CPUs have a number of general-purpose registers (amount and size depending on whether you have a 32- or 64-bit x86 CPU, and the operating mode it's in), a number of floating point registers (or rather, a stack), and a number of multimedia/SIMD registers (also depending on how new the CPU is). Operations available also depends on the type of registers (GP, float, SIMD).