Very Computer

I fired up my Linux again so I could read this newsgroup. I forgot how
to access the drives with KDE3. I think I must have hit everything at
least twice but no A,B,C,D,E,F that I can find. I must be too OScentric.

In comp.os.linux.advocacy, Dennis

 wrote
on Sun, 15 Dec 2002 04:11:10 GMT

Quote:> I fired up my Linux again so I could read this newsgroup. I forgot how
> to access the drives with KDE3. I think I must have hit everything at
> least twice but no A,B,C,D,E,F that I can find. I must be too OScentric.

You've spent way too much time in DOSland or WINWorld.
Come over to Unix Heaven(tm). [*]  :-) At Unix Heaven our
drives can have multiple letters and sensible names.

For example, one method I happen to like of partitioning one's drive
might be something like the following:

/boot     small, but usually /dev/hda1 or /dev/hda2 so LILO can get at it
/         100 megabytes
<swap>    2 times available RAM; this can be put anywhere but usually
          is near the top of the drive for performance
/usr      A large partition, about 2 gigabytes at least; if set up
          properly this can be mounted read-only except when one is
          installing or deinstalling software.  Great for security!
/var      This one on my system is about 500 megabytes and is for
          log files and stuff.
/home     "The rest of the drive".

If one has additional partitions (I have FreeDOS sitting around
backed up at the moment; I need to redo one of my systems)
one can name then things like /c or /d or /dosc or /dosd.  Or
you can put everything under /mnt; for example, one might see
the CD under /mnt/cdrom.

NFS introduces additional flexibility, but I won't go
into details here as it's probably not an issue for you
personally.  However, I like to put NFS mounts under /net.
(The automounter option is probably not that useful unless
one has a large number of machines to manage, although it
is convenient -- when it works; NFS tends to hang when a
machine goes down while another machine is accessing it.
Yes, this is a *feature*, but one has to use it carefully.)

The flip side of all this?  One has to explicitly mount
the drive using 'mount'.  Some workaround exist, such as
'supermount', but the effect is somewhat like old CP/M,
if one remembers that far.  However, it's not usually an
issue since most people won't be mounting hard disks, but
floppies and CD-ROMs.

The exact mapping of physical drive partitions to logical
pathnames is usually specified in /etc/fstab, and the
format is reasonably simple.  Usually the installation
software will go through a number of steps and create
this file for you while it's partitioning the drive(s)
and initializing the file systems -- the rough equivalent
of DOS's FDISK and FORMAT commands.

One gigantic advantage of Unix Heaven is that,
if one adds additional drives, one doesn't have to
worry about the letters shifting around in weird ways.
For example, suppose one does regular backups (always a
good thing, although I'm not sure how many of us do this
regularly enough :-) ) and an IDE drive dies.  IDE drives
(usually) use the nomenclature /dev/hda1 or /dev/hdb2 or
/dev/hdb5 for their partitions.  Now, since Christmas is
approaching one might want to replace that dead IDE with a
high-performance SCSI drive (actually, nowadays SCSI seems
to be getting to be less of an option but I've not checked
lately).  The nomenclature there is (usually) /dev/sda1,
/dev/sdb2, etc.  DOS would have many problems with this,
but in Linux one need merely partition the drive ('fdisk
/dev/sda' -- do this very carefully, BTW, on a drive with
already-mounted partitions), initialize the partitions with
mkfs.ext2 or mke2fs, mount them, and go.  (Or, better,
put the new physical device names in /etc/fstab first,
then do a 'mount -a'.)

In DOSLand one might have a drive with two partitions
on it, which show up as C: and D:.  If one adds another
drive, with two partitions on it, the D: partition
may become E:!  This is sanity?  In Unix Heaven one has
a little more control as to where data is put.

(I'm given to understand that Win2k introduced the concept
of a Partition Manager to attempt to work around the above
problem but have not yet seen it in operation.  Perhaps it
fixes this particular problem.  Of course it, like the
defragmenter, is fixing a problem that ideally wouldn't
be a problem in the first place, in a saner system.)

[*] Linux borrows a large number of concepts from Unix, at least
    at the user level.

--

It's still legal to go .sigless.

Datagram from Dennis incoming on netlink socket

Quote:> I fired up my Linux again so I could read this newsgroup. I forgot how
> to access the drives with KDE3. I think I must have hit everything at
> least twice but no A,B,C,D,E,F that I can find. I must be too OScentric.

You mean drive letters? There are no drive letters in Linux. Every accessable
file (in local namespace) belongs to single filesystem tree. New filesystems
can be attached to it and unnecressary filesystems removed from it on
demand and while the system is operational. With automounters, it can even
happen on the fly (E.g. go to /floppy, and if there is floppy in the drive,
it is mounted automatically. When the floppy is unused, it is unmounted).

-Ilari
--
Guys, Linux is not a research project. Never was, never will be. --
Linus Torvalds
Linux LK_Perkele_IV9 2.4.18 #8 SMP Thu Mar 14 19:42:25 EET 2002 i686 unknown
  5:30pm  up 160 days, 22:59,  6 users,  load average: 0.15, 0.11, 0.05