Coridon Henshaw wrote in a message to Mike Bilow:

 CH>> I'm probably totally out to lunch on this one, but maybe
 CH>> this loophole exists so OS/2 can CHKDSK the boot drive
 CH>> (which can't be locked, correct?) on bootup if required.

 MB> No, the file system has not been mounted at the point when CHKDSK
 MB> would be run.

 CH> I told you I was probably wrong on that one :-)

 MB> I know this seems paradoxical, but the installable file systems are
 MB> loaded quite far along into the boot process.

 CH> So the /AUTOCHECK portion of the IFS=hpfs.ifs line is parsed
 CH> and acted on by some other part of the boot process?  That
 CH> seems quite messy. 

The file system is mounted using a kind of iterative process if the boot
partition is managed by the file system in question.  Things are much more
straightforward if you are not worried about the boot partition.  There
really is no easy way, for obvious reasons, to allow booting from an
installable file system that may be corrupt, and to then perform repairs as
the boot proceeds.

Very few operating systems can do this.  NetWare never boots from a NetWare
partition, but rather is started from DOS.  Unix gets away with ckfs by
using a monolithic kernel, but OS/2 must actually load files from the boot
partition in order to access the boot partition.  OS/2 plays a very
elaborate game of musical chairs to make this work at all.
 
-- Mike


---