Following is a suggestion for a minor change to the filesystem which (I hope)
should make the system slightly more robust.
Traditionaly UNIX uses inode 2 for its root inode, (whereas Linux uses
inode 1, a holdover from minix), and inode 1 is not allocated.
I suspect that this inode (1) was originally used for a bad blocks file,
like .badblocks, which *did not* appear anywhere in the file system.
From what I have seen of System V systems the use of inode 1 in this manner
seems to have ceased and the file system now has its bad blocks remapped
to good ones making the whole filesystem appear to be good.
May I therefore suggest that we follow the original (?) practice, namely use
inode 2 for the root and an unattached inode 1 for the bad block file. This
will allow people with MFM drives (such as myself) to (almost) act as if we
have a perfect drive. We still won't be able to read the device but it will
prevent accidents from happening when the bad block file is accessed or
deleted (mcc interim 96c nearly got me :) ).
It will also incidentally bring us in line with the rest of the UNIX world.
If we are to make this change then, with the extended file system here and
a FFS on the horizon, we should make it soon before the current organisation
becomes too widespread.
Obviously this will also require that fsck and mkfs are updated, but those
are fairly minor changes (which I would be willing to do.)
A utility could be written to allow swapping a pair of inodes would
allow users who don't wish to re-mkfs their discs to move .badblocks to