We are past the stage where the OS has become the virus

[towforce]

Here's desktop OS share in 2016 - link

Here's desktop OS share over the last 12 months - link

* windows share has fallen from the high eighties to the high seventies

* but most of that lost share has gone to OS X - Apple's OS

* Chrome OS is disappointing, having only grown a little (IMO a really good OS for most people)

* Linux is even more disappointing, having barely grown at all

The news is better when you look at other platforms - especially mobile, where Android is the top choice (maybe not the greatest OS ever written, but at least it's not Windows!)

The lesson for me looking at those charts is that choice of OS is basically religion.

The best hope going forward is that the desktop steadily diminishes in importance (you can see this by looking at the "all platforms" charts). The best hopes for prising people away from the grim misery of Windows could be:

1. people find that a "lesser device" than a desktop is "good enough"

2. a cheap desktop which doesn't use Windows proves to be "good enough" (this was the big hope for Chrome OS - but it's only happening slowly)

3. the keyboard becomes obsolete and the concept of the desktop changes (maybe the smart speaker with touch screen is the start of this, and goodness knows why so many smart TVs are still controlled by a TV remote - a device designed to change the volume or the channel of a TV). IBM's PC concept is 40 years old now. When you realise how powerful a cheap SOC is today in comparison to comparable technology from 1981 when the PC was launched, it seems reasonable to ask, "Why are we still using them?"

[flok]

When you realise how powerful a cheap SOC is today in comparison to comparable technology from 1981 when the PC was launched, it seems reasonable to ask, "Why are we still using them?"

Maybe because they're usually less powerful.

A "cheap SOC" can maybe start your browser in for example 5 seconds, but if you then see your neighbour open the browser in a second on "real PC" then I can imagine you want that as well. That's just an example.

[towforce]

A "cheap SOC" can maybe start your browser in for example 5 seconds, but if you then see your neighbour open the browser in a second on "real PC" then I can imagine you want that as well. That's just an example.

Here's a video showing an SOC based device opening a browser - link. I didn't time it, but I don't think it was quite a full five seconds.

[flok]

A "cheap SOC" can maybe start your browser in for example 5 seconds, but if you then see your neighbour open the browser in a second on "real PC" then I can imagine you want that as well. That's just an example.

Here's a video showing an SOC based device opening the internet - link. I didn't time it, but I don't think it was quite a full five seconds.

It was only a theoretical example.

My point was (and is) that they're usually less powerful than a full blown core i7 etc processor.

It'll do for my 4 year old daughter, but eventually she'll become older and maybe then wants to use more performance demanding software (certain games for example) - this also is a theorecital example. Things may go different specifically for her but it may apply to others.

[towforce]

A "cheap SOC" can maybe start your browser in for example 5 seconds, but if you then see your neighbour open the browser in a second on "real PC" then I can imagine you want that as well. That's just an example.

Here's a video showing an SOC based device opening the internet - link. I didn't time it, but I don't think it was quite a full five seconds.

It was only a theoretical example.

My point was (and is) that they're usually less powerful than a full blown core i7 etc processor.

It'll do for my 4 year old daughter, but eventually she'll become older and maybe then wants to use more performance demanding software (certain games for example) - this also is a theorecital example. Things may go different specifically for her but it may apply to others.

The success of the PC 40 years ago was largely because everyone was craving for IBM to set a standard for microcomputers - but is "one device to do everything" still the best option? When you can get an SOC for $0.50 (not much of one by today's standards - but MASSIVELY more powerful than the original PC!), probably not. You probably want different devices for:

* viewing videos

* answering ad-hoc questions or playing music

* spreadsheets/documents

* games

* monitoring your health

* social networking

etc.

Another thing is that Windows is bloaty, coercive and expensive. Most people probably don't mind this on their desktop or laptop - but if a cheap and lean alternative became available that was "good enough" and which had a compelling reason to use it, maybe people would start to move.

[mvanthoor]

Come on. I remember a new user here, Pi4Chess, who was initially asking for Raspi ARM Linux binaries, and who said he wasn't a programmer so that he wouldn't know how to compile. It didn't take many explanations, and guess what, he pretty much compiled everything that would even run on a Raspi.
For someone who is a programmer, that would be even less of an issue.

It's not an issue; I just don't want to spend the time of compiling an engine on top of testing it for stability.

At least not for the vast majority of engines because they are written in C/C++ so that there is a proper compiler already installed by default, and you don't have to download some LTS-distro bypassing, nightly compiler toolchain.

That snipe is unnecessary. The reason why Rust works as it does, outside the distro's, is because the open source and Linux community LOVE doing work 50 times over. Many distributions have different package managers, or even work differently from one another. You can't put Rust into an LTS and then not update it for two-plus years. The language develops so fast, with a release every 6 weeks, that Linux distributions can't keep up. Rust of two years ago is ancient history.

Interweaving the operating system with applications (with everything basically) and putting it all into a single repository is the biggest mistake that was ever made. Yes, it's convenient for updating, but it's gruesome to keep applications up to date, while the OS stays the same.

Windows + a package manager (such as Choco), MacOS with App-installers, Android + APK, FreeBSD + ports is the only correct way; and Linux is finally catching on with Linux + Flatpak. But even there they can't decide on a standard: there's Flatpak, AppImage, and of course Ubuntu must have its own, so they have Snap which the entire world seems to hate.

In the case of Rust, you just install it the same way on every operating system. After you did, it'll have installed rustup (Rust's package manager/updater), and the latest stable toolchain for the platform you're installing on.

[flok]

Interweaving the operating system with applications (with everything basically) and putting it all into a single repository is the biggest mistake that was ever made. Yes, it's convenient for updating, but it's gruesome to keep applications up to date, while the OS stays the same.

Please clarify: I often see updates coming in for things like the chrome browser etc while kernel/glibc/etc updates are way less frequent?

Also managing things as one large consistent thing has the added value that you won't have dll-version problems. Or hundreds of copies of a certain dll scattered throughout the filesystem. Diskspace is cheap, but let's say a security bug is found in version x of a dll. With windows you need to go through your whole system to find each individual dll and check its version. And hope that the new version won't break that application.
When I click on update on my linux system, I can be sure that everything still works afterwards.
Of course I can all of that myself, but the whole idea behind a computer is that things get automated. I see no fun in manually keeping tons of seperate systems and applications up to date. I'd rather let the linux-distribution I run do so. For the things I do, ubuntu/debian and fedora do this pretty well.

I'm not too fond of all those seperate packagemanagement going on. It "weakens" the consistency guarantee of the main package management system.

And if you insist to have the latest version of some program, then that's always possible: just put in your home-directory or so, seperate from the linux-distribution. If that's not possible, then there are alternatives like "containers" and vms and so on. That's more complicated, yes, but that only applies if you insist doing things outside of the linux-distribution.

[towforce]

Looking at the last two posts, there's an opportunity here: a web server that receives source code, a script, a way to automatically discover which OS you're running (e.g. from the web browser), and sends back compiled code.

There are plenty of sites that compile online (link), but not much that makes it easy and comfortable.

One problem you'll get is that some software needs libraries (and Windows' .NET suite is stuffed to the brim with libraries), but this level of convenience should be the ultimate aim, and the website should maximise convenience as much as possible.

[mvanthoor]

Looking at the last two posts, there's an opportunity here: a web server that receives source code, a script, a way to automatically discover which OS you're running (e.g. from the web browser), and sends back compiled code.

In FreeBSD, the Ports system downloads source code, and compiles it locally. In the Linux world, Gentoo is one of the distro's that does this. That case you'll always have code compiled perfectly for your machine, running the fastest way possible. The downside is that installing a large application could take 5 hours.

[Ras]

The best hope going forward is that the desktop steadily diminishes in importance

Which it won't because the decreasing relative importance is due to the growth of smartphones and tablets - but these are devices for consuming content, not creating.

the keyboard becomes obsolete and the concept of the desktop changes (maybe the smart speaker with touch screen is the start of this

It's not like voice assistants had not yet been invented, after all, but smart speakers for an acoustic user interface would make offices pretty awful. Plus that monitors with tons of fingerprints are disgusting. Also, the human ergonomics are exactly the same today as they were last century because our bodies have not changed. People won't want a cramp in either their arm from constantly reaching out to the monitor, or in their neck from looking down on a table integrated screen. Not to mention that many countries have regulations on workspace ergonomics.

When you realise how powerful a cheap SOC is today in comparison to comparable technology from 1981 when the PC was launched, it seems reasonable to ask, "Why are we still using them?"

Because we don't work on 80x25 monochrome text screens with a single task OS anymore.

That snipe is unnecessary.

You argued that compiling was a hassle. All it takes is either clicking a build script, or right-click, open terminal, enter the make line. That's a few seconds more than just having the binary, i.e. nothing compared to stability tests, but also allows the compiler to optimise for your specific CPU. Sure I did notice the irony that Rust is more complicated because the build chain suffers from the inner platform effect.

The language develops so fast, with a release every 6 weeks, that Linux distributions can't keep up. Rust of two years ago is ancient history.

The point of LTS distros is exactly to escape from software with such release cycles and the usual fallout from that. RR distros can keep up - just use Arch, and your versions will be so current that their source code is still warm.

Yes, it's convenient for updating

And also for security because directly downloading binaries from all your software vendors has its own set of problems. Solarwinds anyone? That will take years to clean up.

The downside is that installing a large application could take 5 hours.

Yeah, for LibreOffice or Chromium, that wouldn't be too nice, but for chess engines, it's quite feasible, and chess engines are speed hungry applications. Only that this is already easy locally if you use an OS that usually comes with a compiler already pre-installed.