[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Partitioning

Bill Leach <bleach@bellsouth.net> wrote:

 > In scsi (and I suspect IDE), the drive responds to the block
 > address and "knows nothing" about partitions, logical or otherwise (or
 > for that matter, clusters).

  IDE drives (at least, the older ones) are addressed in cylinder, head
and sector numbers, but you are right, partitions are logical divisions
of the drive that the drive itself know nothing about.

 > Though it depends upon what one might mean by the term "knows", the PC
 > BIOS _IS_ the implementation of this particular filesystem abstraction.

  You constantly confuse the issue.  The BIOS 'knows' how to load a
sector from disk to memory or vice versa.  This has _nothing_ to do with
filesystems.  Filesystems define how data is ordered and stored on these

 > The decisions as to how to deal with the initial design limitations from
 > the original and even subsequent BIOS designs is what has "given" us this
 > "nightmare" design.

  I'm not exactly sure what you think is the nightmare part of the
original design (and frankly, I don't care).  There are a couple of
limitations that we have to live with, given the fact that the PC was
designed in days where a harddisk of 10 MB was way to expensive to
consider, and that manufacturers still want to maintain backwards
compatibility with software written in those days.  A few of these are:

 * The standard MS-DOS MBR software is not capable of addressing
harddisk sectors as logical block numbers (i.e., it uses a cylinder /
head / sector addressing scheme).  Modern BIOSes are capable of using
sequential block numbers, though, and e.g. W95 _will_ install MBR
software that uses these (and install itself on a partition with a type
that MS-DOS doesn't understand) if installed on a HD/partition of above
(I believe) 2 GB.

 * There is a limited number of primary partitions available in the MBR.
This limitation is no serious problem, as many modern OSes don't object
to being installed in an extended partition (of which there can be as
many as required).  Of course the MS-DOS MBR software does not support
booting them, but should modern PC hardware be judged by old software?
There are plenty decent boot mangers around.

 * The IDE hardware interface limitations, combined with the older BIOS
interfaces, limited the addressable (through BIOS) size of a harddisk
to 504 MB.  As early as '92 the first workarounds for this problem
appeared; they are (confusingly) known as Large or LBA mode, and
basically work by translating cylinder, head and sector numbers in such
a way that the limit is increased to 2 GB.  Meanwhile, newer BIOS
interfaces are present on all modern BIOSes that don't even have this
limitation, and modern OSes are capable to use these interfaces.

 > Linux, for very practical reasons, chooses to honor this brain dead,
 > convoluted drive abstraction.

  There is nothing brain-dead about partitioning a drive - it is a
perfectly logical way of having multiple independent filesystems and/or
operating systems on one disk.

 > As to lilo -- IT IS lilo that has the problem.  While under Linux, lilo
 > determines and stores the block locations for the files that it must
 > access to boot the machine, when the boot occurs those blocks must be
 > accessible through BIOS calls.  It is the limitations of the BIOS calls
 > that started this whole mess!

  Not in modern BIOSes.  It is backward compatibility that makes LILO
use the old-style BIOS calls, but it is probably (by now) capable of
making the newer style calls if told to (I can't check that as I don't
have the latest version here).  (Note that it is also backward
compatibility that made the PC to what it is now: powerful hardware for
very low prices).

 > I repeat:  There is _nothing_ inherent in Linux that requires any of this
 > "grew like topsy", screwball "design".

  Ignoring your qualifications here, I never said there was.  I was
objecting to your misinformation about the basic PC hardware and BIOS.

 > As can no doubt be discerned, I detest the basic PC design.

  I have no desire to change your opinions on the basic PC design, but
if you want to comment on it (and especially if you start answering
other peoples questions), at least get your facts straight.

 > I would be
 > willing to defer to your claims that I am wrong concerning primary
 > partitions being visible in DOS, but am having some trouble with that.
 > My experience with DOS and Windoz is _very_ limited.

  Mine is quite extensive.  Why do you not believe me, when you yourself
say that your experience here is _very_ limited, and I am telling you I
have the very thing you claim to be impossible running on my system (and
on quite a few others, I might add)?  Why do you combine lack of
experience with such strong opinions?


Gertjan Klein <gklein@xs4all.nl>
The Boot Control home page: http://www.xs4all.nl/~gklein/bcpage.html

TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-user-request@lists.debian.org . 
Trouble?  e-mail to templin@bucknell.edu .

Reply to: