98SE
Really, it is ALL written up ALREADY in this thread, please RE-READ (and re-read again, and again the related posts) NO value higher than 0xFFFFFFFF is possible in any of the 2 fields related to LBA addressing (Start Address and Size).
There is NO space for any bigger value.
Imagine that you have a cabinet with 8 drawers, each drawer can either contain one single object (and no more than one).
How many objects can you store in the cabinet?
8 and no more than 8.
Now you have a field 8 characters in length, how many characters can you store in this 8 characters field?
8 and no more than 8.
Since allowed characters are 0123456789ABCDEF, knowing that F represents the bigger possible number, what is the biggest you can write?
FFFFFFFF and no more than FFFFFFFF
Now these are hex numbers, you can use Windows calculator to play with them just fine, 0xFFFFFFFF (hex) is 4,294,967,295 (decimal).
What happens if you sum 1 to 0xFFFFFFFF?
Here:
0xFFFFFFFF+0x1=0x100000000
you need 9 characters, as the 8 you have available will be 00000000.
Sure , this is what we (highly specialized tehnicians) call "logical volumes inside Extended Partition" and exists since DOS 3.2 or so.
The only issue here is that the Extended Partition "contains" the logical volumes, so it must be bigger that the sum of all volumes inside it, with a maximum size of - guess it - 0xFFFFFFFF.
Would such an Extended Partition with a "fake" (limited to 0xFFFFFFFF size) work nonetheless through the chain of EMBR's (notwithstanding the name RLoew used, this is another thing, the term EMBR is commonly used to indicate the Extended MBR's that allow indexing logical volumes inside the Extended partition).
So the Extended has values:
0x00000800 0xFFFFFFFF <- i.e. it begins at offset 2048 and extends for 4,294,967,295 sectors
The values for the first volume in the EMBR at 0x00000100 are as well:
0x00000800 0xFFFFFFFF<- i.e. it begins at 2048+2048=4096 and extends for 4,294,967,295 sectors
Now where would the next volume begin? At 2048+2048+4,294,967,295=4,294,971,391, i.e. 0x100000FFF that - once written in the space available will become 0x00000FFF, or 4,095, neatly beginning before the first volume.
Sure, it is just a matter of writing a driver (a kernel one will be needed for having the disk bootable) capable of accessing the whole disk RAW and create an on-the-fly structure virtually combining two or more extents in a single virtual volume.
The hint about a (super-)floppy (that you didn't seemingly catch) by cdob is actually an interesting one, however. and could well be the target for a further experiment.
Most probably a filter driver (such as the reversed dummydisk,sys) to make the hard disk a "Removable" (please read as "non-partitioned") device might be needed (or maybe not) for the test on an internal disk.
The test on a USB super-floppy makes little sense since (if) the USB adapter already translates sector size to 4 Kb, there are no issues up to 17.6 Tb, thus it sounds more than anything else a solution in search of a problem.
It would be needed an USB adapter capable of accessing >2.2Tb disks while NOT translating sector size, which looks even more pointless.
jaclaz