OSDev.org

Hey!

Well, I haven't been around this part of the internet for a while. In any event, I have a discussion topic that may have already been instigated at some point or another, but I'm looking for present thoughts on the subject. I recently read something about the number of lines in the Linux source code doubling over the course of 36 months. What with the everbearing nudges about the bloated nature of Windows, I can't help but wonder about the possibility of these both becoming obsolete in the near future. If so, what would the future of operating systems look like? Considering the time it actually takes hobby open-source operating systems to reach a certain level of functionality, I see a possibility of open-source operating systems becoming obsolete. That is at least in the sense of the Linux kernel being widely community developed. I doubt the likelihood of another OS being able to take the same development path. In the event of widespread obsoletion of present-day popular operating systems, I see that propriety development would have the advantage of being able to fund the development of an entirely new modern OS within a reasonable time-span. In that case, do you see any significant design changes kernel architecture?

nekros wrote:I recently read something about the number of lines in the Linux source code doubling over the course of 36 months.

I am not an expert but I think that adding more drivers to it does not make the actual kernel more bloated. As far as I know, drivers are the biggest part of it. Of course Linux (and GNU, Unix-like) has lost its fun a long time ago. It is time to have something new.

Kernel design is not rocket science now.
The situation is that you aim at a goal (or usage) and that affect the kernel design/model - this in turn differentiate your kernel/product.

Now, the challenging part is not designing a kernel, but pick the correct goal(s); the kernel design follows naturally once you have the goal.

nekros wrote:do you see any significant design changes kernel architecture?

Yes, because there are new goals.

For example, traditionally, the GUI is tightly coupled to the OS and it's in many way affect the OS design - IMO the GUI is part of the OS.
The GUI was then assume there is keyboard, mouse, and WIMP - this started to change when people getting used to touch-based device, and the introduction of touch screen notebooks, or integration of UI with AR...so the GUI design changed.

For internals like process management, the hardware changed to many cores, huge address space, etc; while the usage changed from "just hundreds of processes, a few active applications" to the trend of "millions of processes, hundred of smart agents/applications"- this can affect kernel design as well.

But before you invest zillion dollars for new OS, history told us that the feature sets, or coolness of an OS alone do not make it success, the most successful OS is merely a mess in any sense but with some good marketing, and extremely lucky timing to hit the market.

Antti wrote:

nekros wrote:I recently read something about the number of lines in the Linux source code doubling over the course of 36 months.

I am not an expert but I think that adding more drivers to it does not make the actual kernel more bloated. As far as I know, drivers are the biggest part of it. Of course Linux (and GNU, Unix-like) has lost its fun a long time ago. It is time to have something new.

I read somewhere that Linus feared the day when Linux becomes to large for developers to work with and troubleshoot. I could be wrong.

I remember Linus complaining about Linux being bloated 4 years ago (that article was published on my birthday, heh). The truth is, it's difficult to have a huge project not become bloated sooner or later.

The other thing is that many people like to talk about how software is bloated but they don't really know what they mean by that---just that it's Something Bad(TM). As a test, I propose that the OP describe how Windows is bloated (after all, they reduced the hardware requirements and improved the performance significantly with Windows 7---haven't used Windows 8 so far). I don't say that in a fastidious manner; I wouldn't know how to answer that question because I don't know enough about Windows' internals.

My intention in this post, btw, is not to advocate one OS over another.

nekros wrote:I can't help but wonder about the possibility of these both becoming obsolete in the near future.

Why would they? I don't see this happening anytime soon.

For the sake of argument, let's assume that they are indeed bloated. This bloat happened not just because someone felt like writing bloated code, but because there's an awful lot of features and functionality that are useful for people. Which of course means that these people won't switch to a different OS that can't provide the same functionality, just because it is said to be less bloated.

Considering the time it actually takes hobby open-source operating systems to reach a certain level of functionality, I see a possibility of open-source operating systems becoming obsolete.

Erm, and keeping the source secret makes functionality appear magically all by itself?

I see that propriety development would have the advantage of being able to fund the development of an entirely new modern OS within a reasonable time-span.

You clearly haven't understood open source and that free speech/free beer thing.

Open or not, source needs to be written. If you want a functional modern OS developed in a reasonable timespan, closing the source isn't going to make it any easier. It just means you have to recruit programmers differently.

It also means you won't be getting patches from outside of your team, nor will bug reports be as useful.

If you look at most operating systems out there, many of them are decades old and they all use the same basic primitives. Linux is from the 90s, Windows NT from the the 90s, QNX from the beginning of the 80s and these use designs usually based on much older operating systems. I would say that things are moving quite slow anyway. I don't see paradigm shift anywhere but I haven't searched that thoroughly though.

I think the paradigm shift will come with new hardware and not the other way around.

bluemoon wrote:while the usage changed from "just hundreds of processes, a few active applications" to the trend of "millions of processes, hundred of smart agents/applications"

In desktop use, I do not think this is true. Most people only need few active applications. Of course threads are going to be used more efficiently. "Hundred of smart agents/applications" sounds like an OEM customized Windows installation. Those "smart" agents with annoying popups are one thing which make PC "less cool" for most people.

I read somewhere that Microsoft is quite disappointed that OEM bloat software makes Windows less attractive. I truly believe this point.

nekros wrote:
I am not an expert but I think that adding more drivers to it does not make the actual kernel more bloated. As far as I know, drivers are the biggest part of it. Of course Linux (and GNU, Unix-like) has lost its fun a long time ago. It is time to have something new.

I read somewhere that Linus feared the day when Linux becomes to large for developers to work with and troubleshoot. I could be wrong.

And rightly so; If bloat is said to be software with 'too many features', Who is to say it's too many? - For me its the developer, and 'too many' is when they can no longer manage the code for them. [1]http://bertrandmeyer.com/2012/11/27/why-so-many-features/

Antti wrote:I read somewhere that Microsoft is quite disappointed that OEM bloat software makes Windows less attractive. I truly believe this point.

Half of these bloated software packages are usually updaters for thing like Adobe Reader, Java, hardware drivers, and the like. I think Microsoft should put serious thought into opening up Windows Update to other software providers. So when you click "Update" it figures out that you have version 7.10 of Java and that Oracle has released 7.11 through Windows Update, and then goes and downloads and installs it for you. Much better to have one single reliable updater software than one good one and a hundred crap ones floating around on the system.

IMHO, this is the killer feature on GNU/Linux, and the thing that most GNU/Linux distributions do better than every other operating system out there. You run 'aptitude update && aptitude upgrade' and it figures out what updates are available for every single piece of software on the system. The stores (Windows Store, Google Play, iTunes Store, etc) are really a more complex package management system (where you have to pay for the right to install some packages) - and they make a lot of sense. Surely a lot better than the alternative of downloading unsigned installers off the Internet, which then install the software but also a web browser toolbar that steals your credit card and a badly-written custom updater than chews 10% of the CPU constantly.

Thinking about it, I guess the Windows Store is the beginnings of an all-encompassing Windows Update. Could be a good thing as far as OEM bloat is concerned; time will tell.

</$0.02>

JackScott wrote:Much better to have one single reliable updater software than one good one and a hundred crap ones floating around on the system.

That is true. I do not think it would be technically very difficult to have an API (provided by Microsoft) for using Windows builtin update mechanism. Or is it possible already? As far as I know, MSI packages have quite a lot of features.

In addition to my previous post, the main problem definitely is the user space. For example, Windows 7 seems to be quite OK with clean installation. The scheduler or memory management (all "kernel things") are good enough and it makes that aspect irrelevant for normal users. In desktop use, I do not think alternative OSs can compete in that field. To give an example of kernel insignificance: Linux could be replaced and still keep the "Linux feel." If Ubuntu changed Linux to Hurd/BSD/Antti/whatever, it would look exactly the same "Linux" to normal users.

The user space program "administration" is the thing that matters nowadays. What I am trying to say is that kernels itself may not "have any future." If hardware is infinitely fast and we have infinite memory (e.g. 256-bit memory space?), do we need any special handling for them? An extreme example would be "malloc with no need for free and hundreds of cores per a process." Kernel could be even simpler than today (in theory).

What I am trying to say is that kernels itself may not "have any future." If hardware is infinitely fast and we have infinite memory (e.g. 256-bit memory space?), do we need any special handling for them? An extreme example would be "malloc with no need for free and hundreds of cores per a process." Kernel could be even simpler than today (in theory).

l
I think we all view a kernel as abstraction from the underline hardware, so while there is different hardware, there will always be a kernel.

I personally see the kernel as a Controller object. It would be a facade controller if a uniform interface was all that was needed, but i suspect a small amount of logic will always be required.

cxzuk wrote:so while there is different hardware, there will always be a kernel

This is also a good topic. I feel that in the future OSs do not need to support very wide range of hardware. Perhaps we finally get "universal" interface at hardware level? To add more, perhaps the functionality mechanism is available on firmware? Then the actual OS just takes care of policy. Another thing is whether "PC-like" systems where hardware and software are from different vendors are going to last.

As an historical side note: when GNU* project started, they made user space programs first. After then, it was time for kernel. For example, there are some members here who have a high goal of making their operating system to success. However, they do this in reverse order. Although, here in this forum we concentrate more on kernel so I cannot know their priorities for sure. It does not matter in which order to make things if everything is done perfectly. However, my current opinion is that user space things have more important role in achieving the ambitious goal.

*(GNU is not a very good example because it had a reference OS: Unix.)