@Karlos

Quote:

Karlos wrote: @cdimauro A monolithic kernel is only more complex when it's more complex. What do I mean by this? If the design isn't "right" you'll have problems. A monolithic kernel that tries to include everything and the kitchen sink will run into the problems you describe. A monolithic kernel that concerns itself with only the most fundamental aspects of memory, threads, hardware initialisation and delegates everything else into cleanly segregated userland processes doesn't have to end up in dependency hell.

Then this isn't a monolithic kernel, rather an hybrid kernel.

BTW Linux is monolithic (the fact that it's modular doesn't mean that it's like a microkernel).
Quote:

The same observation is true for microkernels. If you take the idea to extremes and partition services too far and too granular you end up with Hurd and all the emergent complexity that arises from what seems like a simple design at first.

I don't think so. QNX, again, proves that it's working very well.

@cdimauro

What is a hybrid kernel, really? It's just the name we assign to something that is somewhere between the two concepts of monolithic and microkernel. I think by that measure, all modern successful kernels are hybrid, it's just a question or how close they are to one ideal or the other. It's about as useful as RISC and CISC these days (ok, that's probably not s great comparison as there are more factors to consider here, but still).

QNX and Linux are successful in their respective "micro" and "monolith" corners because they aren't monodimensionally obsessed with the label. Hurd is unsuccessful because it got too far into the idealism of what it is to be a microkernel at the expense of pragmatism.

Last edited by Karlos on 24-May-2022 at 09:58 PM.

_________________
Doing stupid things for fun...

@agami

If you dislike Linux so very, very much… why you use it?! Masochist?

Last edited by kolla on 24-May-2022 at 11:14 PM.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

Quote:

kolla wrote: @agami If you dislike Linux so very, very much… why you use it?! Masochist?

Hmm, it might play a little into my mild martyr complex.

I do not dislike it "so very, very much". I just simply dislike it. But I dislike macOS and Windows even more.
It is a tool for a job, and because I have had the privilege of using really good tools, I can see that LInux is not as good a tool for the job, but it'll have to do for now.

Just because something is the least worst option, does not make it a good option.

_________________
All the way, with 68k

@Karlos

Quote:

Karlos wrote: @agami Quote: When I say I have been dissatisfied and continue to be dissatisfied with Linux, I mean everything about it, e.g. I'm not a fan of the monolithic kernel Objectively speaking, what is it that you dislike about the monolithic kernel? A monolithic kernel doesn't have to be huge. A monolithic kernel simplifies a lot of design aspects that become trickier in microkernel architectures. The fact that the GNU/Hurd team still don't have a properly working microkernel solution for their own userland and the Linux kernel was able to fill that gap, despite being monolithic does lay bare some hubris. Subjectively speaking, as a "cog in the machine", how often does it's monolithic nature impinge negatively on your everyday experience of using a Linux distribution?

I was illustrating the point when I say I dislike a platform, I take into account my personal userland experience, the experience of other groups, and my philosophical and architectural principals.

Don't Build Monoliths, happens to be one of those principles.


Quote:

Architecturally, and purely subjectively, I do prefer microkernels but it's not a hill I'd die on.

Then what is the proverbial hill to die on?

If you subjectively prefer a microkernel architecture then there has to be some good reasoning behind it. Why diminish your subjective view in the face of some fictional objective view? What are objective views if not just a collection of subjective views? More often than not emerged out of convenience. There is no Cosmic preference for how we humans construct our labor saving devices.

The enemy of the Excellent is the Good Enough. I happily charge this hill every day.

_________________
All the way, with 68k

@Karlos

Quote:

Karlos wrote: @cdimauro What is a hybrid kernel, really? It's just the name we assign to something that is somewhere between the two concepts of monolithic and microkernel. I think by that measure, all modern successful kernels are hybrid,

I don't think so. QNX, again, is an example. Zircon might be another one.
Quote:

it's just a question or how close they are to one ideal or the other. It's about as useful as RISC and CISC these days (ok, that's probably not s great comparison as there are more factors to consider here, but still).

Well, here I think very differently because almost all processors are CISCs. According to the definition of RISC (see Patterson etc.).
Quote:

QNX and Linux are successful in their respective "micro" and "monolith" corners because they aren't monodimensionally obsessed with the label. Hurd is unsuccessful because it got too far into the idealism of what it is to be a microkernel at the expense of pragmatism.

From the QNX wiki page:

"The QNX kernel, procnto, contains only CPU scheduling, interprocess communication, interrupt redirection and timers."

Isn't it small enough to declare it a ("pure"?) microkernel?

@kolla

Quote:

kolla wrote: @agami If you dislike Linux so very, very much… why you use it?! Masochist?

Mostly because I'm forced to use it at work. Or because there's no other suitable solution (at home).

I share similar experience and opinion with agami here.

@agami

So in your world, you only have three options when it comes to operating systems, and all of them are bad?

You poor thing. Maybe you should just create your own operating system …

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@cdimauro

QNX may be the perfect microkernel example. Shame it got swallowed up by blackberry. I'd happily run a GNU userland on it otherwise.

My point about kernels is ultimately that only a diminishingly small number of users care about whether it's monolithic or not, as long as their devices work. People like us are not representative, even remotely, or the consumer market for computing devices.

@agami

Let's say you built the perfect OS. Secure. Scalable. Performant. As close to universally cross platform as is feasible. Do you think people will flock from Apple, Android, Windows, etc? Most people I know are extremely resistant to change. People using windows because they've always used it. Or Apple. The only real mobility seems to be in the phone/tablet space.

_________________
Doing stupid things for fun...

@Karlos

Quote:

Karlos wrote: @cdimauro A monolithic kernel is only more complex when it's more complex. What do I mean by this? If the design isn't "right" you'll have problems. A monolithic kernel that tries to include everything and the kitchen sink will run into the problems you describe. A monolithic kernel that concerns itself with only the most fundamental aspects of memory, threads, hardware initialisation and delegates everything else into cleanly segregated userland processes doesn't have to end up in dependency hell. The same observation is true for microkernels. If you take the idea to extremes and partition services too far and too granular you end up with Hurd and all the emergent complexity that arises from what seems like a simple design at first.

A monolithic kernel has a huge benefit in that internal APIs do not need to be set into stone
and it allows you to iterate and improve, updating designs over time.
The kernel is self-contained and no internal APIs are exposed out to "external" components.

This makes for example full blown API rewrites like the current Linux page management and filesystem datahandling to switch to a completely new model : folios, possible.

If you don't have a monolithic kernel you end up in situations where you can not do these kernel-wide api and abstraction redesigns because you might not be able to coordinate these changes across all modules in an atomic way. Maybe you end up with versioned APIs and modules that translate between different api versions? lol nightmare. Maybe you need to build compatibility layers and emulate different versions of the APIs?

Monolithic kernels have drawbacks but they have huge benefits in maintenance and innovation velocity because you CAN rip out and completely replace one internal API with a completely different one without having to worry about internal-api-compatibility with the previous but now obsolete abstraction.

Last edited by V8 on 25-May-2022 at 10:33 AM.

@V8

to me the discussion is rather academic in general. For a platform to attract users it must be easy useable and offering a look&feel they like. Much of it is desktop related. Also important is what software is available and how good everything is integrated.

if the kernel is monolithic or not is not interesting. More important is what advantages or drawbacks a choice offers. Would a complete new system with a new kernel but drawback not being compatible to anything and thus no drivers and software really offering feelable advantages that justifies this?

This discussion is much too technical and not looking at it from a user point of view. But users decide about success or failure. You can of course do your private hobby OS that only is used by yourself but at least for me that would not be very interesting.

For me if someone really thinks about new he or she should think what are the features and advantages it should offer to the user and then discuss what to use and not vice versa.

Last edited by OlafS25 on 25-May-2022 at 10:50 AM.
Last edited by OlafS25 on 25-May-2022 at 10:48 AM.

@kolla

Quote:

kolla wrote: @agami So in your world, you only have three options when it comes to operating systems, and all of them are bad? You poor thing. Maybe you should just create your own operating system …

That is one of my goals.

_________________
All the way, with 68k

@agami

assuming someone would install your completely new OS what would be the advantages from a user point of view? It would be incompatible to everything so you would need new software and drivers for it as I understand it. What would be the advantages compared to a linux or windows system?

@OlafS25

As I said it's not something almost the entire consumer market thinks about. If they do think about their OS today they think in terms of what they need. Stability, security, performance and ease of use. The finer technical details of how those facets work they don't care about. OSX is not the most technically excellent OS (I'm not a true hater, it's not a bad OS, just not for me), but for the most part, sans the occasional passwordless root login faux pas, it delivers those facets to the user who frankly just wants to get on with doing other stuff.

_________________
Doing stupid things for fun...

For me, something like AmigaOS is the perfect hobbyist OS. I understand it's limitations. Like those crazy guys who make gas fuelled pulse jet trikes. They know their machines will never be roadworthy transport and don't fret about it.

_________________
Doing stupid things for fun...

@Karlos

Quote:

Karlos wrote: @agami Let's say you built the perfect OS. Secure. Scalable. Performant. As close to universally cross platform as is feasible. Do you think people will flock from Apple, Android, Windows, etc? Most people I know are extremely resistant to change. People using windows because they've always used it. Or Apple. The only real mobility seems to be in the phone/tablet space.

Well for one, in order for it to be "perfect" it would make switching from other platforms to the new one satisfying experience.

I don't need users to flock. 5% in 5 years is the aim.
Many Windows users have in frustration tried to move to macOS or Linux, but have returned to Windows because they find it the least frustrating.
Many macOS users have their grievances with Apple and the direction the OS in which Apple is taking the OS, but they mostly stick around for the apps.
There are many Linux users, much like myself, who know what they would like Linux to be, and they know that Windows or macOS could never be.

And then there are the dissatisfied developers on all of these platforms.

In order to have any chance at success, one can't come at these head on: Ta-da, here's your new perfect OS!
A flanking maneuver is required.

_________________
All the way, with 68k

@agami

but then you must first understand why those users are frustrated. Is it desktop/configuration so your workflow you are using up to now no longer works and they have learn new? Is there software missing they used?

If there special apps that are only available on one platform and not run on yours you can do what you want but people will not change to it. And to improve Linux experience, what are the reasons for frustration. I assume in many cases a different desktop and configuration would do the job.

@OlafS25

Quote:

OlafS25 wrote: @agami assuming someone would install your completely new OS what would be the advantages from a user point of view? It would be incompatible to everything so you would need new software and drivers for it as I understand it. What would be the advantages compared to a linux or windows system?

First, it cannot be incompatible with everything. I would never work to deliver such a product. Bridges are important.

Hmm... let's see. What's the best way to put it without writing a treatise:
- For developers, developing software in half the time of any other platform
- For users, near-precognitive levels of user interactivity. Knowing what the user wants to achieve, and making it easier to achieve than on traditional "legacy" platforms.

_________________
All the way, with 68k

@agami

OK... then you would first need to analyse the typical workflows on other platforms and do something different. But also offering compatiblity here (from a user point of view) so that a user can still do it the way he is used to from f.e. windows. Users are lazy typical and dislike to learn everything new even if it is more efficient finally. One reason certainly why a windows users seldom changes to MacOS as a example (besides of missing software). And also automation is important because f.e. MacOS professionals have highly automated the workflow.

@OlafS25

Quote:

OlafS25 wrote: @agami but then you must first understand why those users are frustrated. Is it desktop/configuration so your workflow you are using up to now no longer works and they have learn new? Is there software missing they used? If there special apps that are only available on one platform and not run on yours you can do what you want but people will not change to it. And to improve Linux experience, what are the reasons for frustration. I assume in many cases a different desktop and configuration would do the job.

I have all the necessary supporting data.
This isn't some idea that sprung into my head just the other week. The first seeds of this go back to about 2003, and it didn't get fully formed until about 2013.

I know what you're thinking, in all that time how come I haven't done anything about it? Like it's that easy.
Execution is the key to its success. I've been busy with starting other businesses, in order to have the revenue to support the initial work required before bringing on investors.

_________________
All the way, with 68k

@agami

No I do not think it is that easy. In reality it is a huge challenge to create and establish a new platform on todays market. You already have different options that are well-known with lots of support, many users, huge software bases. That is the reason why I am always skeptical about creating something completely new. I think from a user point of view many problems are not OS or hardware related but depend on desktops/user interface and how the system is configured. So to create a new desktop with better user experience using a established system looks more realistic to me. But it is not my project but yours.

Last edited by OlafS25 on 25-May-2022 at 12:22 PM.
Last edited by OlafS25 on 25-May-2022 at 12:21 PM.