> I know most cp/m utils write to a drive blind and generaly abort if the drive > fills up but the OS does allow apps to tell the drive-space/size just apps > don't bother looking! Plus cp/m flat directory structure allows files to > traverse many drives, again apps just don't bother using the facility due to > the extra coding needed to configure such apps to the different drive > characteristics of the thousands of different drive setups.
Yeah - I hate the way MS-DOS does that. It's not hard to look at the contents of the FAT to work out the number of free clusters, and *then* start to write the file, not the other way around. Some people, eh? > Cs> I'm considering allowing a > Cs> linked-partition system, but not sure > Cs> about how it'd perform speed-wise. > Cs> Basically you can define two E-DOS > Cs> partitions and link the data space > Cs> together to form one big > Cs> partition. It *may* work -- gotta look into it more. > > Sounds complicated! Are you designing the EDOS or whatever to be a system > thats designed for the user to have consious knowledge of what the thing is > doing or more a virtualised system that isolates the user from everything > involved? It'll be more virtualised -- if you link paritions (in the final DOS), then they will appear to be the same -- you will be able to allocate partitions as subdirectories and vice versa. I'm thinking of implementing disk images -- you copy a SAM disk in, and tell the DOS to map it to drive 1 or drive 2 -- then you can do all the sector access as normal :) > Cs> I'm considering using the > Cs> logical-sector access mode of the IDE drive > Cs> to store all stuff -- other people for > Cs> other partitions can use the > Cs> sector/cylinder/head method if they > Cs> want :) This should speed up > Cs> access -- you just take the sector > Cs> number in the partition, add it to > Cs> the logical start of the partition and > Cs> read / write... quick 'n' easy :) > > Hey, I missed that! Is that a logical translation feature of the low-level > ATA-HD protocol or an interface layer in the host driver software? > > If it's the former it'll simplify the banked-uzi port I have in mind as > that's > designed around logical sector-sized blocks on a scsi HD:-) It's the former -- LBA mode, set as part of the features register, allows you to use 24bit (I think -- not got the specs in front of me) logical sector numbers to address the hard disk. Will make it a *LOT* easier, and also will make it a piece of cake to convert to SCSI or even to an IDE CD Rom drive using its seconds based address scheme. Si Cooke
