From: cross@spitfire.i.gajendra.net
In article ,
Waldek Hebisch wrote:
>Dan Cross wrote:
>> In article ,
>> Waldek Hebisch wrote:
>>>Dan Cross wrote:
>>>> In article <87o6ybbeqw.fsf@example.com>,
>>>> Salvador Mirzo wrote:
>>>>>scott@slp53.sl.home (Scott Lurndal) writes:
>>>>>
>>>>>> "Paul Edwards" writes:
>>>>>>>Sure - but why not make it available anyway?
>>>>>>
>>>>>> MS-DOS is, was, and always will be a toy. It's not even
>>>>>> a real operating system.
>>>>>
>>>>>And why is that? Is it mainly because it doesn't time-share the CPU?
>>>>
>>>> It depends on your definition of an operating system, I suppose.
>>>> I like the definition Mothy Roscoe (ETH) used in his OSDI'21
>>>> keynote:
>>>>
>>>> The operating system is that body of software that:
>>>> 1. Multiplexes the machine's hardware resources
>>>> 2. Abstracts the hardware platform
>>>> 3. Protects software princples from each other
>>>> (using the hardare)
>>>
>>>This is oversimplified definition, any definition of similar
>>>length will be oversimplified. But let us see how this
>>>works.
>>
>> Mmm...not really.
>>
>>>> It's hard to see how MS-DOS fits that definition in a meaningful
>>>> way. Does it multiplex the machine's hardware resources? Well,
>>>> no; not really.
>>>
>>>[snip]
>>>
>>>Also, you seem to ignore a file system. For definition
>>
>> Funny how in the very next paragraph you quoted, I was talking
>> about a filesystem. ;-P
>>
>>>above to make any sense multipling machine hardware
>>>resources must include mutiplexing (coordinating) access
>>>to external storage which is (part of) function of file system.
>>>
>>>> While it does provide a primitive filesystem,
>> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
>> (See note above)
>
>You did not state relation with "multiplex the machine's hardware"
Actually, I did. Perhaps you are having a hard time
understanding what I wrote? Is there some way I could make it
claerer?
>and quick "Well, no; not really" suggests that you do not
>count this as multiplexing, I think that you should.
See above.
>>>> and exposes some interface for memory management, it only lets
>>>> one program run at a time, and that program doesn't have to use
>>>> or honor DOS's filesystem or memory management stuff. Further,
>>>> the system interface is inexorably tied to the hardware; it's
>>>> defined in terms of synchronous software traps and specific
>>>> register values. System calls are numbered, not named.
>>>
>>>System calls are numbered in almost all operating systems.
>>
>> You're talking about the ABI.
>
>Yes, that is what matters for programs.
>
>>>[snip]
>>>> Finally, the last one is really the nail in the coffin: MS-DOS
>>>> makes absolutely no effort to protect the software principles
>>>> from each other, or even themselves; a user program can take
>>>> over and just never cede control back to DOS.
>>>
>>>Well, DOS is close to best possible protection given the
>>>hardware. In modern times hardware protection gained
>>>importance, but putting hardware protection as mandatory
>>>part of operating system definition distorts history
>>>quite a lot.
>>
>> By the time the IBM PC came along, we'd had systems where the
>> OS was protected from errant programs for 20 years. For example
>> look up the Manchester Atlas system.
>
>Sure, there were system with memory protection. But a lot
>of hardware had not memory protection and even now such
>hardware is in extensive use (granted some foljs are not
>aware of them and even more would not count them as computers).
>
>>>There is a lot of valid critique of DOS, but saying that it is
>>>not an OS is just silly game of words. You can pile adjectives
>>>on OS, like "multitasking OS", "proteded OS" (or better
>>>"OS using hardware protection") and DOS will be outside such
>>>restricted classes of OS-es. But is clearly an OS.
>>
>> Well, except perhaps it is not. At least not by a very
>> reasonable definition that's widely accepted in the field.
>>
>> I really don't see why people are so upset about this; it's not
>> a huge deal. DOS was ok for it's time and for what it enabled
>> on the original IBM PC; the hardware was very limited, and so it
>> wasn't nearly as capable as larger systems with "real" operating
>> systems. Why is it a priori a bad thing to acknowledge that?
>
>I do not care about DOS. And I acknowledge limitations of DOS.
>I do care about clear terminology. Terminology where removing
>memory protection from Linux (to make it run on hardware not
>capable of memory protection) turns it into "not an OS" is
>a nonsense. Concerning widely accepted: I do not think that your
>interpretation of definition is widely accepted. I mean,
>putting hardware memory protection as part of definition
>may acknowlege its importance for some operationg systems,
>but leave it optional in general. If hardware memory
>protection was intended as mandatory thing, then this is
>politcal statement with which I think a lot of specialists
>disagree.
As I said, this is the definition due to Mothy Roscoe at ETH.
It was given in the keynote for one of the two major conferences
on the subject. Perhaps watch for yourself and then judge.
https://www.usenix.org/conference/osdi21/presentation/fri-keynote
>For example Wirth gives "system Oberon" which he
>calls operating system, but which has no hardware memory
>protection.
Speaking of ETH....
>More generally, basic terminology should be incusive. It
>is easy to add extra qualifiers to narrow meaning. It
>is awkward to use phrases "something like OS, but which
>does not satify some random guy definition of OS".
Sounds like you have some studying to do, my boy.
- Dan C.
--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)
|
