<on disk size supported (say number of 256-byte logical sectors), a more
<robust (economical, speedy) disk structure could be developed.
<
<If we allowed a 32-bit logical sector pointer, then 2^32*2^8 bytes could
<be stored. That is 2^40 bytes total. Since 2^20 bytes is 1 Meg, a 32-bit
<logical sector pointer allows 1 megamegabyte. 2^30 bytes is 1 Gig, and
<2^10 is 1024. Therefore, 2^40 bytes is 1024 Gig.
<
<I think a 32-bit pointer would be sufficient for a while. Most drives
<available are under 20 Gig.
Several key items.
Most more recient disks are 512byte sectors.
Even with 16bit pointer using clustering 512k is not unreasonable, though
maybe inefficient.
Using a 24bit pointer and 512b sectors will net 8gb, adaquate.
The problem (pardon my lack of knowledge of the TI Filesystem) is that
simply having a point of some size is only part of the problem. You also
need to have a descriptor for how many sectors are owned by the file in
question. If the file system is a simple sequential one an additional
two or three bytes are adaquate. Generally though a sequential allocation
scheme is inefficient for random file access. A random file access scheme
means multiple descriptors for each block (assuming multipe scetor
granularty). Obviously in that case you can really consume space in the
directory with a desicriptor per block. This leads to FAT or ALLOCATION
style of file systesm (dos and cpm respectively) or something like unix
uses. I'm suggesting that more bits extend greater addressing but impose
other impacts not mentioned.
Allison
Received on Wed Jan 13 1999 - 20:33:27 GMT