En SpaceButler va escriure:
The directory structure is oriented to clusters. What happens if some
great partitioning utility has the ability for the user to set 64K or
128K clusters? Is there any reason scandisk would not work with
those?
I guess 16-bit Scandisk will have problem to load a 64K (or worse a 128K)
cluster in memory; my recollection is that Scandisk does not use the "huge"
model for that (since the performance hit would be dramatic).
Since it is required to do that to analyse at least the folder structure, I
would think this will hang.
Several similar tools will have similar problems.
OTOH, as far as I know one can disable Scandisk at boot
(inside MSDOS.INI, AutoScan=0).
Also, once you avoid Scandisk at boot and use only 32-bit tools, I do not
see why the "traditional" setup with 32 KiB or even 16 KiB clusters, and
hence huge (>16 MiB) FAT, could be a problem...
Also see below your experiment.
By the way, the 16 MiB limit is purely a matter of compilers used (which
memory manager targetted the 80286 and where not updated). Using other
schemes while still using 16-bit code, one can go upwards until 256 MiB
without too much of a problem (256 MiB is a limit for two reasons: one is
that it correspond to 64Ki pages of 4KiB each, and also it fills half of the
LDT, the part which is not used for the shared descriptors.)
My guess is that except internal tests at MS, nobody will.
If it does work, then the size of a partition can increase to 255.06
GB or even 510.12 GB. This would be perfect for large drives storing
large files.
Perhaps. I would be happy to hear about experimentations here. But I am not
holding my breath.
I don't think one should put the OS on a partition like that.
Surely not! the code which is loading DOS then VMM and the VxDs is in very
wide part still 16-bit code, and much of it is dealing with directories and
cluster content, and it definitively would not like having to deal with
clusters bigger than its internal buffers (something as easy as seeking
IO.SYS in the root directory is already a problem.)
You would still be limited to a Win98 max file size of 4GB.
I seem to remember there is a limit around 2 GiB (signedness).
Could be wrong, though.
For the fun of it, I did format the 300GB drive as one large
partition. (I forgot the cluster size, though the fat was huge.)
Win98 seemed to store files properly and could read them,
Once you have booted, Windows is running with 32 bits code, which does not
have any problem of any kind (except at 2/4 GiB, but this is one was solved
long ago.)
There may be a partition table limitation.
Yes, but it is quite a bit higher. The limit is using 32-bit numbers for
sectors, so this translates to 2048 GiB or 2 TiB. Note this limit is inside
the partition table (MBR), but it is also replicated in several places, such
as in the boot record for the file system, so it is quite of a hard limit
for FAT.
It is worth an experiment.
It would cost a bit (e.g., with four 500GB disks in hardware RAID0/JBOD you
are still below the mark), but this can be tested yet. It raises a number of
funny problems, BTW. I have no idea how Win9x could react... nor am I rich
enough to give it a try.
Antoine