Subject: mtools, demand-loading, paging
Date: Thu, 14 Nov 1991 13:22:37 +0200
From: Linus Benedict Torvalds <torvalds@cc.helsinki.fi>
To: Linux-activists@joker.cs.hut.fi

Hi again.  Several people have inquired about mtools.  The sources I
used are available on nic.funet.fi: pub/unix/386ix/isc/source/mtools*. 
I don't know if this is the most recent version.  And no - linux
probably cannot read partitions >64MB.  The reason is (partly) the minix
file system.  See later. 

Paging to disk has also come up: It will certainly be implemented (in
fact one of the bigger changes in the 0.10 fs was due to making ready
for this). There are no obvious problems with this, but unless somebody
else wants to try to implement it, it will take some time (I'm talking
March-92 at the earliest, probably not until summer). One of my problems
is (will be) lack of disk-space to try it out on.

As for demand-loaded executables... hmm. There are a couple of problems
with this: the minix filesystem wasn't meant for it, and it would be
MUCH easier to implement if BLOCK_SIZE was 4096 (= one page) instead of
the current 1024. This of course means waste of disk-space. The problem
with the minix filesystem isn't that serious: I'll have to move on to
something else anyway - 64M partitions can be limiting for those who have
disks big enough to hold them (and the afore-mentioned 64Mb limit even
for direct read/writes.. This I can try to do something about).

The more serious problem about demand-loading is that the fs isn't
really designed for it - I doubt the locking mechanism used currently
will suffice. If people would accept a non-locking demand-loading (ie
you could write to the file while it's in use etc...) I could possibly
try to implement it. Again - not until well into -92.

And a little small sad note: I know I said symlinks would probably be in
the next version, but unless I can get the higher priorities completed,
this will be left for later.  The things now relatively certain (barring
acts of God etc) are: fsck, mkfs, badblocks, fdisk, non-VGA video and
finally the correction to the serial drivers to change speed by the
ioctl call. I'll look into extended partitions (fdisk will probably
report them), but I doubt they will be accepted by the kernel proper
until later.

		Linus

PS. I try to answer all mail as soon as I get it, but I get MUCH mail.
If you think the mailing-list has been active: think again :-). If you
don't hear from me, I've probably forgotten to answer at once, and once
your mail is in my mailbox, it's likely to be lost among the multitude.
Re-mail your question, and make me feel guilty about not replying the
first time: that's the ticket.