What do you dislike about linux the most?

So I guess the closest thing to a kernel from how you describe it would be DOS's command.com?
 
No command.com is a shell. Closest thing would be /bin/sh. I know that it is hard grasp because command.com is the thing people interact with directly. It makes one think that it is the OS but all it does is basically take the name of a program to execute and then execute it.

There is no kernel in an OS like DOS it does not need one - it does not provide the types of services that would require one.

Philosophically a kernel is about providing a virtual environment for each process and managing these virtual environments. Now DOS is a real mode operating system and does not provide any from of virtualization. There is only ever one process (no preemptive multi-tasking by definition), one address space (no virtual memory, no address translation), IO is done by direct manipulation (no virtual devices). Because these sorts of features do not exist in DOS there can not be any part of DOS which is the kernel providing them.

::existential sigh::

-Eric
 
Hello lurk,

Originally posted by lurk
There is no kernel in an OS like DOS it does not need one - it does not provide the types of services that would require one.

Philosophically a kernel is about providing a virtual environment for each process and managing these virtual environments. Now DOS is a real mode operating system and does not provide any from of virtualization. There is only ever one process (no preemptive multi-tasking by definition), one address space (no virtual memory, no address translation), IO is done by direct manipulation (no virtual devices). Because these sorts of features do not exist in DOS there can not be any part of DOS which is the kernel providing them.

you are right, as you can see http://whatis.techtarget.com/definition/0,289893,sid9_gci212439,00.html

but DOS offer the services:
a process scheduler: only one process at time (not counting TSR)
memory address space manager: full memory access
disk management: file system primitives (FAT12 FAT16...)
service requests are accomplished by software interrupts

OK, it's not a great kernel but it offers minimal services to run shell (command.com) and to execute programs without access disk directly, check the allocated and free memory, etc...
DOS kernel is a very simple kernel for a simple OS offering minimal services.

CBM Commodore 64 kernal offers tape and disk "virtualization" services, memory bounds check... very minimal services enough to work with pheriperals (OK, it's more like a BIOS but that is).

This talk is reaching philosophical arguments so I think that it's time for me to stop. You are right, me too from another point of view so there's no reason to continue.

Byez_
 
Originally posted by binaryDigit

OpenBSD - For those worried about security. Also with multiple platform support, but usually not as good as NetBSD in this regard.

Why wouldn't everyone want a secure system?

Don't you think security should be a requirement in all systems?

If systems came configured for maximum security, we'd all be better off. Users would have to enable the items they needed, instead of systems being configured by default with everything enabled.

I use a router, NAT, and a firewall. I also use WEP and MAC restrictions on my wireless system.

People should demand better security from their systems, no matter what OS they use.
 
Originally posted by Easter
you are right, as you can see http://whatis.techtarget.com/definition/0,289893,sid9_gci212439,00.html

There needs to be a word for the way that a word looses its relevance as a google keyword when it is overused in other contexts. I did the same search you did when I posted my replay but I disregarded that definition because it was too fluffly. It talks around the issue in an effort not to offend beginner ears and in the end says nothing. Any other potential definitions are lost in a slurry of "kernel traffic" and other similar useless hits.

I am afraid that I will have to direct you to the local library for a good definition. The book we used when I TA'ed for Operating Systems was the one at http://os-book.com and it covered the topic quite nicely. This is the classic Dinosaur Book for those of you who have been through a CS program in recent history.

but DOS offer the services:
a process scheduler: only one process at time (not counting TSR)
Think about the logic of what you are saying. No action is performed so what can you say about the mechanism by which it wasn't performed.

It is all or nothing there is no scheduling going on so there can be no part of DOS to attribute the scheduling to.

As to TSR programs they are called "Terminate and Stay Resident" for a reason they are not actively scheduled. They hide out in memory and wait to be called as an interrupt handler or via something like a jump table.

memory address space manager: full memory access
Again total lack of management is not management.

disk management: file system primitives (FAT12 FAT16...)
As I mentioned above IO is not one of the things which makes a kernel and there are kernels where IO is specifically moved into user space processes.

This is not evidence for a kernel.

service requests are accomplished by software interrupts
Yes that was part of the reason I said that DOS was just an interrupt handler. (It is important to note that using interrupts to communicate with the kernel in an OS is a common thing since that can allow transfer of control from one trust level or process to another. So handling interrupts in and of itself is not a bad thing.)

OK, it's not a great kernel but it offers minimal services to run shell (command.com) and to execute programs without access disk directly, check the allocated and free memory, etc...
DOS kernel is a very simple kernel for a simple OS offering minimal services.
But it is not a kernel this is a simple question of definition. This is exactly like trying to argue that a fish is really not a fish but a primitive mammal. I mean it gives birth to live young (eggs are alive are they not) and it has hair (well scales but that is just like hair). Words have definitions for a reason and trying extend the definition of mammal to include fish you remove the usefulness of the terms themselves.

CBM Commodore 64 kernal offers tape and disk "virtualization" services, memory bounds check... very minimal services enough to work with pheriperals (OK, it's more like a BIOS but that is).
Again no kernel here. That is not a bad thing it doesn't mean that some how this is a lesser OS. Edx, if you are in this thread I wonder if an OS can have "Kernel Envy"?

This talk is reaching philosophical arguments so I think that it's time for me to stop. You are right, me too from another point of view so there's no reason to continue.
The problem is that this is not a point of view question again it is a simple question of definition. Accuracy in language is important in very concrete term, the assertion you are making that DOS has a kernel is patently false. It is the same as my assertion that a tuna is a mammal, if we accept both of those in the interest of validating some relativist perception of truth - discourse is dead.

-Eric
 
Originally posted by dafuser
Why wouldn't everyone want a secure system?

Don't you think security should be a requirement in all systems?

If systems came configured for maximum security, we'd all be better off. Users would have to enable the items they needed, instead of systems being configured by default with everything enabled.

I use a router, NAT, and a firewall. I also use WEP and MAC restrictions on my wireless system.

People should demand better security from their systems, no matter what OS they use.

Well as with anything else, it's a tradeoff. I use NetBSD often because it supports a lot of the weird hardware I tend to have around. For my primary machine(s), other things factor in, like availability of software, stability, hardware support, footprint, etc. OpenBSD places their focus on security, and if it has all the software you need to run, then by all means run it. Unfortunately it does not rank at the top in all the other requirements, so while it has it's place (I have it on two machines right now), it does not fit every purpose.

But I get what you're saying though and generally agree. It would be nice if everyone took a serious look at security, but the realities of software development make doing this at the level needed to make really secure systems is beyond the scope of many companies (for both economic and technical reasons).

However one should never be lulled into thinking that they are 100% secure, not matter what your software vendor tells you. OpenBSD's record is great, but primarily because out of the box, many services are disabled. If you are running an email server or samba or openssl, then all these packages have had security issues in the last month alone. So once you try to get some real work done, you start running into all the weak links in the security chain.
 
Originally posted by binaryDigit

But I get what you're saying though and generally agree. It would be nice if everyone took a serious look at security, but the realities of software development make doing this at the level needed to make really secure systems is beyond the scope of many companies (for both economic and technical reasons).

I'm not sure companies can't put better security in their software if they really wanted to. I'm sure the primary reason is the "good old bottom line." It would cost more to produce better code.

The most common security issue with software seems to be buffer overflows. They have been exploited for years. I think the programmers are careless if they don't take the time to prevent buffer overflows in their code.

If software companies would be held responsible for their poor code, and it affected their "bottom line, you better believe they would produce better software.

While any software can have problems, the open source stuff seems to be faster at putting out fixes.

Slammer and Code Red wouldn't still be around if people would take the responsibility to secure their systems, especially when the vendor has the patches available.

Like you say, no software is perfect, and security is a constantly moving target. What's secure today could well be the next buffer overflow victim tomorrow.
 
Erhm, to get back on topic (DOS is dead, btw...):

1.) That its recent development aims at people who don't understand the first thing about computers.

2.) That Red Hat installer now tries (and fails) to recognise my hardware.

3.) That it gets all the public interest. Mac OS X should get that. It's the UNIX for the rest of us (them).

;)

Okay... I'm an elitist asshole, as someone familiar would put it. Linux is a great (and free as in beer and open source) operating systems for creating small-scale (and big-scale) servers and nice (and cheap) X86 workstations. But you'll have to either know something about computers and linux, or you'll have to choose Red Hat and always click on 'secure' when that option is there. And then you'll need to register for their update service, or all your steps at learning linux might be watched by other people™. Okay, the last part may sound a bit paranoid. Turns out, it's also true.

I don't believe in linux on the desktop. However, I do believe in linux for appliances. And one appliance could even be a desktop or notebook computer that comes with a preinstalled, secure linux system that offers everything a user might need and is automatically updated via a broadband connection. But that would take away a lot of linux' appeal, as half of the fun of having linux on your desktop/notebook is to customise it down to the bones (I mean the shell prompt. And compiling your own kernel. But it's like with people who roll their own cigarettes. They also have beards and are a bit rough sometimes. Linux people who roll their own kernels often lack a beard, but they're also tough guys who know their boxes and systems.)

Remember what happened to the internet when it became user-friendly? Yep, all the losers came 'online'. If anyone who can spell 'warez' - ;-) - can also get Photoshop for free at more than 100 Kb/s, I understand that AOL is truly the root of all piracy. They should blame themselves for all those copies of songs they didn't want to spread faster than the proverbial 'fame'. Okay, I'm totally off-topic now, so I stop this long post already.
 
Back
Top