[Cialug] 10yr future of Linux and open source

[Ron]

Nicolai wrote on 2025-08-28 12:30:

> It's fun to write down predictions and a good learning experience to
> revisit them later. What are your predictions for Linux, open source,
> and the tech world in the year 2035?

Oohhh, interesting. And there could not be a more turbulent time for 
making predictions.



> I think "Linux on the desktop" will continue growing and at a slightly
> elevated pace, maybe even reaching 10% marketshare by 2035.

Until recently, I was skeptical. However, in another window at this very 
moment I am looking at a story about:

 > China’s KylinSoft has delivered a major update to its flagship Linux,
 > which Beijing hailed as a great leap forward for the nation’s ambition
 > to develop operating systems that match and exceed the capabilities of
 > western products.

https://www.theregister.com/2025/08/29/kylinos_11/


My current thinking is that it's not just China that is looking to 
reduce its reliance on US technologies, it's *every former ally* as well.

This could put significant downward pressure on desktop Windows for the 
foreseeable future.

Trust, once shattered, is going to take a long time to rebuild (not 
repair).  So - Linux on desktop is suddenly looking more promising than 
I've ever expected.



> I think systemd has won and more distros will adopt it, and some that
> don't will come to an end.

Any Linux distro that hasn't adopted systemd probably exists solely in 
reaction to systemd and will never adopt it.

These are, IMHO, an ever dwindling group of zealots that are aging / 
have aged out of relevance in tech, and eventually there's be virtually 
no Linux without systemd.


> Developers will increasingly write code
> dependent on systemd, being therefore less portable, requiring other
> operating systems to provide expensive workarounds. This will harm the
> ethos of open source making it a little more superficial (a general
> trend unfortunately).

Hot take: systemd works pretty well, and something with some form of API 
compatibility may arise in the BSD world, easing work loads on devs / 
maintainers and providing some benefits of services management across 
all the *nix OSs.

If anyone hasn't seen Benno Rice's "The Tragedy of systemd" at 
linuxconf.au from ~2019, it's worth a watch (he's a FreeBSD dev and author):

Video plus actually insightful discussion:

https://www.reddit.com/r/linux/comments/aot1ga/benno_rice_the_tragedy_of_systemd_linuxconfau_2019/

YouTube:

https://www.youtube.com/watch?v=o_AIw9bGogo



> Currently, we're on the verge of getting 192 core AMD Epyc CPUs running
> at up to 3.7GHz. What will be available in 2035? What if this is a
> standard high-end consumer workstation in 2035, with high-end servers
> doing kilo-core computing? What will people do with unlimited power?

I don't think there'll be that kind of compute power in the average home 
or office.  Maybe for locally run AI applications but as the use of 
these settles into segments where it's useful, they'll also become more 
efficient and easily run on a phone or laptop.


I guess it's like, no supersonic passenger jets exist and "no one" has a 
car that travels > 200mph parked in their garage.


> I used to think ARM64 would be the next
> dominant platform but now I think it will mainly pressure AMD/Intel to
> advance PPW (performance per watt).

Agreed, I don't foresee many high-core computers in homes / offices but 
lower wattage computing with somewhat higher performance than now.


> So maybe open-source RISC-V could be
> in 2035 where ARM was around 2020, with the potential of being the next
> great thing. Hope so but I doubt it, as RISC-V seems also to be building
> a culture of tinkering with SBCs. That's cool but not for 99.9% of
> computer users.

Again, the shift away from US controlled tech (US gov being the largest 
owner of shares of Intel makes Intel a dubious purchase like Huawei is 
by western telecom companies - at the direction of their governments).

This seems promising for RISC, as ARM has significant US patents I believe?



> I want to believe Ladybird will grow and succeed.https://ladybird.org

I wish I could be hopeful here, but I've heard / read repeatedly that 
the number of pages comprising the specs for HTML, CSS, JS, and related 
technologies used in the web space number about a million. (I cannot 
find any links to back that up, and asking ChatGPT gives "5000-10000 or 
more, depends on lots of things".)


We need Firefox as the sole stronghold against the Chrome / Chromium 
tsunami.  As a non-Apple user, Safari will always be safely ignored 
(perhaps at my peril, don't care).


> Memory safe PLs will continue to make agonizingly slow progress. I hope
> "Rust in Linux" will try again. I think memory safe languages will need
> to take over userspace before operating systems make significant moves
> in the kernel. But that seems to be slowly happening.

I've recently stumbled into using the `fish` shell as a replacement for 
`bash`.

It's designed for enhanced interactive experience first, and sane 
scripting second.

Written in Rust, discarding the anachronisms and cruft of `bash`, it's 
amazing.

Unexpectedly, I think I may also start using fish's scripting language - 
I can read it, the man pages have *examples*, it's not some inscrutable 
punctuation soup.



As new, younger people take up careers in tech, the peculiarities of C 
and bash, etc. are going to be replaced by more modern paradigms in more 
modern languages.

It's a good way to make one's mark and they won't have the "it was hard 
for us, it should be hard for you" mind-set ("we walked up hill both 
ways, in the snow").

So, user space tool replacement is a solid prediction.


Side note: why did my love/hate relationship with bash tip over to "hate"?

Consider the exit codes, crucial if running with `set -e`:

((1+1)); echo $?
0
((-1-1)); echo $?
0
((-1+1)); echo $?
1

Since 0==success and 1==error, the last operation exit code is failure.

But no, when math inside ((...)) has a result of zero, the exit code is 
non-zero for success.  WTF? Go home `bash`, you're drunk.

Dave from YSAP always has excellent bash videos:

 > The Problem with Bash 'Strict Mode'
 > Why I Don’t Use `set -euo  pipefail`
 >
 > https://www.youtube.com/watch?v=4Jo3Ml53kvc


Oh, in fish, with its floating point math:

$ math -1+1; echo $status
0
0
$ math -1/2; echo $status
-0.5
0




> Thoughts? Go ahead, share your own 2035 hopes and predictions.
> 
> Nicolai

Thanks for your predictions and the thought-provoking exercise.