What

(section 1)

Firstly, what is an operating system really. Well, it's a piece of software, nowadays usually written in a mixture of C/C++ and assembly language. Part of it is loaded into the computer's memory when the machine boots and stays there til the machine powers down; other parts may be loaded and unloaded into and from memory as required. Typically after initialising data structures and querying I/O devices, the OS will load an application program called a 'shell' (or User Interface). The shell is thereafter responsible for interacting with the user and asking the OS to load other programs on behalf of the user.

definitions of 'operating system'

How can we pin down exactly what we mean by 'operating system'? (section 1.1)

  • Extended machine, or virtual machine. Takes the basic hardware, and provides additional layers of functionality. The virtual machine is usually simpler and more consistent than the real machine. Practically, additional layers will build on the lowest-level ones, providing a rich programming model for the applications software (which in turn provides a 'virtual machine' to the user).
  • Resource manager. Allocates scarce resources, like disk space, memory, processor between competing processes and users. Each process thinks that it has its own machine (its own files, its own printer, whatever). In a way, this is also extending the machine (in order to appear to be many machines).
  • Anything which runs in the computer's privileged mode, or supervisor mode. Typically most modern processors have two or more operating modes, user and supervisor. In supervisor mode, able to access all of memory and all hardware.
  • The software which is supplied with the machine. Often includes many things not traditionally considered part of the operating system, like GUIs, editors, Web browsers, and stuff.
  • A platform on which other software can run. Lots of platforms are not operating systems (like Netscape Navigator, or Microsoft Word). Java is almost an operating system. Provides many services commonly associated with Operating Systems, like managing multiple concurrent processes, and providing access to files.

    • Why

    Why build operating systems? Why install one on your computer?

    Because a computer, even a very simple one is, like, vastly complicated. If you want to do anything useful, it's reasonable to use existing applications programs rather than writing the software yourself. Similarly, even if you do write your own software, it is more reasonable to make use of other software instead of writing everything from scratch. End users and applications software alike rely on other software performing the lower-level tasks.

    What is there in the naked machine (unclothed by an OS)?

    That's it. You don't have files. All you can see of the disks and other devices is a wodge of hardware registers. You do everything to devices by setting device registers and responding to interrupts. When an event occurs, the CPU receives an interrupt which must be dealt with (and if you aren't careful, you can lose track of where you were in the program). Memory is just in one (or several) big contiguous block(s), and may not be enough to run the applications you want.

    just how tricky is it to program the naked machine?

    You (the application programmer) could deal with all this complexity yourself (and quite recently many games programmers did), but consider the simple task of reading some bytes out of a file on a floppy disk (section 1.1.1):

    The book goes into some detail about the NEC PD765 controller chip on the IBM-PC and how READ and WRITE commands require some 13 parameters packed into 9 bytes, specifying a block, number of sectors and what to do with a deleted-data-address-mark---which return 23 error and status fields packed into 7 bytes; how you can't leave the motor on for too long, but you don't want to keep switching it on and off...

    On my machine at home, you still have to decode the data from the flux encoding used on the physical disk. Then there's still the problem of locating data when reading and allocating storage space when writing (making sure that new data does not overwrite old data).

    The point is that it is tricky, and any one of several things could go wrong. How much better if someone could write code to deal with the general case, which could be reused by anyone wanting to use the device.

    This is what the operating system does.

    last updated 6 February 1997