ALL BLOG POSTS AND COMMENTS COPYRIGHT (C) 2003-2020 VOX DAY. ALL RIGHTS RESERVED. REPRODUCTION WITHOUT WRITTEN PERMISSION IS EXPRESSLY PROHIBITED.

Sunday, November 10, 2019

Linux converged

As expected, once Linus Torvald permitted a code of conduct to be installed, the convergence of the Linux Foundation didn't take long. It's now against the Linux code of conduct to a) wear a MAGA hat, and b) take a picture in front of Trump Tower.


It's time to start cracking down hard on SJWs and methodically excising them from your organizations. Zero tolerance is the only viable policy, as there is literally nothing they will not hesitate to ruin for the flimsiest of reasons.

Also, stop joining converged organizations! What is the point? They're just going to cancel you as soon as they realize you are not part of the hivemind.

Labels: ,

142 Comments:

Blogger ThatWouldBeTelling November 10, 2019 5:55 AM  

If http://techrights.org/wiki/index.php/Linux_Foundation is to be believed, besides funding about a dozen people's real work including Linus, the Foundation doesn't do much that's really worthwhile with an 100 million per year budget.

One of the best pieces of hard evidence of its convergence is this page for individual donations: https://www.linuxfoundation.org/about/donate/

"100% of donations received go towards funding diversity programs."

Blogger Rabid Ratel November 10, 2019 5:56 AM  

Time to fork the linux kernel. I have been getting some weird bugs lately, so it may be best to regress to a kernel before 5.0.0

Blogger FrankNorman November 10, 2019 5:58 AM  

Now I'm wondering how soon all this convergence will result in crappy code in the kernel, because excluding it would have given some snowflake a feelbad.

Blogger Dave Dave November 10, 2019 6:01 AM  

We just fucking wanted to use a cool operating system.

Blogger ThatWouldBeTelling November 10, 2019 6:05 AM  

My posting with a bunch of links was accepted by Blogspot but then removed by a spam filter. Unless and until the author is canceled, go to https://old.reddit.com/user/BornInTu and see the most recent of his set of links with a bit of framing and commentary. Lots of craziness.

Blogger Shimshon November 10, 2019 6:06 AM  

unauthorized.tech - Home of the unconverged Linux kernel.

Coming soon?

Blogger Rabid Ratel November 10, 2019 6:08 AM  

@FrankNorman, I am running linux kernel 5.2 with Gnome 3.34 on an AMD system. Since upgrading to kernel 5.0 I have been experiencing freezing, slow file operation which lock the system until the file operation is finished a la DOS, and plenty of crashehes I have never experienced before.

I am most definitely getting all the benefits from the snowflakes, and not just in the kernel

Blogger Rakshasa November 10, 2019 6:11 AM  

Knew linux was fucked the moment a trans weirdo was allowed to write the CoC.

And before that there were those two programmers at a conference that got fired after a female 'programmer' found their jokes about merging and forking repositories to be sexual harassment, as a 3rd party.

Why do we even let these wamen into our space? They can't code for shit, disrupt everything and create drama always.

When the linux kernel conferences try to get a gender diverse panel going all they can find is someone who commits spelling error fixes and minor code changes once a year in order to claim the title of 'linux kernel dev'.

Blogger Rabid Ratel November 10, 2019 6:11 AM  

@Shimson, there are plenty of techies right here on this blog. Why wait? Let us make a Dark Lord OS that blows Linux out of the water.

Blogger bramley says "Enoch was right" November 10, 2019 6:13 AM  

Make benevolent dictators great again!

Blogger The Cooler November 10, 2019 6:15 AM  

It's as if SJWs occupy an entirely distinct spatiotemporal plane.

Blogger ThatWouldBeTelling November 10, 2019 6:18 AM  

@4 Dave Dave:

"We just fucking wanted to use a cool operating system."

I recommend OpenBSD: https://www.openbsd.org/ It's Benevolent Dictator For Life (BDFL) makes most of them like the old Linus look weak, it's uncompromising on its openness and security goals, it does what its developers want, which includes running well on a set of laptops. One very important thing, a rare continuation of the original UNIX, is that manual pages are comprehensive and complete, changes to the system are not accepted without updating them.

Major compromise for the average user is that for browsing, for now you'll probably want to use their spin of Chromium, which they're hardening against mischief in their own ways. Another is that it updates every 6 months, only 1 year support for a version, but the updates are pretty gentle as long as you're willing to do a bit of work in modifying a config file or two. And I think they're working on making that even easier.

Avoid FreeBSD, it got converged some time ago, and that's showing in quality issues, and for example abandoning its fork of ZFS for the Linux one.

Blogger Paulito November 10, 2019 6:19 AM  

What do hats have to do with operating systems?

Blogger bramley says "Enoch was right" November 10, 2019 6:24 AM  

Slave mentality people who get upset by the term 'slave' used in a technical context.

CreepyFreaks won't be happy until everything is broken but 'safe'.

Picture the future as a robot endlessly walking into a wall with a rainbow flag on it.

Blogger bramley says "Enoch was right" November 10, 2019 6:31 AM  

Time to start using https://infogalactic.com/info/TempleOS

Blogger ThatWouldBeTelling November 10, 2019 6:33 AM  

@8 Rakshasa:

"Knew linux was fucked the moment a trans weirdo was allowed to write the CoC."

The very most cancerous one, (((Coraline Ada Ehmke))) https://en.wikipedia.org/wiki/Coraline_Ada_Ehmke (don sunglasses before going to that page), who's now pushing a "Hippocratic Licence" which forbids users of your software from doing evil as he sees it.

"And before that there were those two programmers at a conference that got fired after a female 'programmer' found their jokes about merging and forking repositories to be sexual harassment, as a 3rd party."

That was Donglegate, at a conference for the Python programming language, which to a large degree is killing itself from bad technical management, most especially the breaking changes from version 2 to 3. That had perhaps the very best outcome of any of these, the SJW who got 1-2 developers fired turned out to be an evangelist her company, that is, someone who's job is to convince the same developers to use their stuff, and genuinely became unemployable.

"Why do we even let these wamen into our space? They can't code for shit, disrupt everything and create drama always."

Airport's Law is one explanation: "Every day another goony beard-man gets the impression that a rainbow haired she-twink might let him cum in her if he attacks gamergate."

One bottom line from the subset of these events that are at conferences is that they are to be avoided at almost all costs, the substance is watered down by the required inclusion of wamen and other minorities, the opportunities for networking are steadily smaller, and have to be weighed against the potential for getting canceled from your career.

Blogger Player 2 November 10, 2019 7:15 AM  

Everyone best grab a current version of their favourite flavour and set it to never update again. Now that they're converged, it's only a matter of time before it degrades to the point of being unusable. I also have a couple Live versions on USB that I think I'll keep in case they eventually force downloaded versions to update.

Blogger SmockMan November 10, 2019 7:17 AM  

Karhu wrote:@8 Rakshasa:

"Knew linux was fucked the moment a trans weirdo was allowed to write the CoC."

The very most cancerous one, (((Coraline Ada Ehmke))) https://en.wikipedia.org/wiki/Coraline_Ada_Ehmke (don sunglasses before going to that page),


It has the thousand CoC stare.

Paulito wrote:What do hats have to do with operating systems?

When they next attack Red Hat Linux for being a symbol of hate.

Blogger Shimshon November 10, 2019 7:41 AM  

@9 @Shimson, there are plenty of techies right here on this blog. Why wait? Let us make a Dark Lord OS that blows Linux out of the water.

Rabid Ratel, I would probably contribute to such a project. Starting a site, forking the kernel, and hosting it, are not really sufficient. Successfully recruiting high level kernel hackers tired of the CoC BS would be a boost.

Is anyone seriously interested in this? It strikes me as a pretty big undertaking, and this is the time and place to speak up.

Blogger FrankNorman November 10, 2019 8:15 AM  

Is anyone seriously interested in this? It strikes me as a pretty big undertaking, and this is the time and place to speak up.

I don't know what I could do to help - kernel-level programming is way out of my league.

Blogger Rakshasa November 10, 2019 8:17 AM  

Caring about the kernel is so naughties, today everything is about containers.

The time will come when we have to fork docker, not any one kernel.

Blogger Crunchy Cachalot November 10, 2019 8:30 AM  

Gotta love that the torpid she-cow's chimp-out, replete with F-bombs, is considered "professional" by the Linux Foundation, indeed, is viewed as an irrefutable indictment.

Of course, it was stupid and weak of the accused to attempt to dialogue with ideologues.

Blogger Rabid Ratel November 10, 2019 8:31 AM  

Shimson, I am definitely interested.

Blogger Ron Winkleheimer November 10, 2019 8:36 AM  

Of course, it was stupid and weak of the accused to attempt to dialogue with ideologues.

The guy is still under the delusion that reason can win the day. Reason only works with reasonable people. If they could SJWs would gleefully put dunce hats on us, pelt us with dung, and then send us out to die in the countryside working on collectivized farms. That's a hard thing to internalize for some people.

Blogger Damelon Brinn November 10, 2019 8:36 AM  

@19, I would be more inclined to invest my development or anti-convergence efforts in one of the BSDs. Linux has been making poor technical choices like systemd for a while, whether because of the growing convergence or just parallel with it, I don't know. I started moving to BSD about 20 years ago because it seemed like Linux was trying too hard to compete with Windows by being all things for all people rather than a solid Unix OS. I still use it for some things and admin it for clients, but I use a BSD (usually FreeBSD) when I can.

The main advantage Linux has is a large number of developers to write all the hardware drivers you need to support every desktop and laptop (and various other devices) on the market. It has all those developers thanks to corporate money supporting them, and from converged corporations will come converged developers. If you want an unconverged OS, you have to accept that it won't necessarily run perfectly on the new laptop you just picked up. You have to check hardware compatibility lists or buy from someone who knows.

There are desktop BSDs, but I can't recommend any because I'm old and stuck in my ways, so I still use FreeBSD on mine. FreeBSD's convergence has been somewhat exaggerated. There's been pushback there and a gradual awakening to the threat. Their Coraline, Randi Harper, was kicked out in a rare case of a CoC being used successfully against an SJW. Some of her bootlickers remained, and a couple years ago they got the white knights to put a new CoC in place that's so bad some important developers left the project. I haven't heard much about it since, and it seemed like everyone was tired of the drama and just wanted to get back to work so they agreed to ignore it. If FreeBSD were a cute furry animal, I'd put it in the "threatened" category, not "endangered."

But there's also OpenBSD, which has Theo, if you prefer the one-tyrant defense against convergence. Of course, that's what Linux users used to count on Linus for, but Theo is on another level.

Blogger Ron Winkleheimer November 10, 2019 8:39 AM  

When they next attack Red Hat Linux for being a symbol of hate.

Can I get a link to that? I tried searching the web, but didn't see anything.

Blogger James Dixon November 10, 2019 9:10 AM  

> Linux has been making poor technical choices like systemd for a while...

Red Hat is responsible for that. Poettering is a Red Hat employee.

> Can I get a link to that?

Their logo is a red fedora. That's all it will take. People have already started reacting to simple red hats as if they were MAGA hats.

Blogger MichaelJMaier November 10, 2019 9:21 AM  

This comment has been removed by the author.

Blogger Crew November 10, 2019 9:22 AM  

The thing to do here is to create lots of pictures of Linux Luminaries wearing MAGA hats and flood the internet!

Then complain to the conference organizers ...

Blogger The Lab Manager November 10, 2019 9:23 AM  

So a sheboon named Kim Crayton is upset that a white European male is wearing a MAGA hat? I wonder what her coding skills are and what she has ever contributed to a Linux kernal or any project for that matter? Sounds like she should have been banned. Damn shame we can't ban people like her from the rest of society.

Blogger Ominous Cowherd November 10, 2019 9:32 AM  

Rabid Ratel wrote:@Shimson, there are plenty of techies right here on this blog. Why wait? Let us make a Dark Lord OS that blows Linux out of the water.

Can we have a pre-systemd fork?

Blogger James Dixon November 10, 2019 9:40 AM  

> Can we have a pre-systemd fork?

Slackware and Devuan don't use systemd. Either would make a good base to start from.

Blogger IAMSpartacus0000 November 10, 2019 9:40 AM  

Bloods are for Trump. Who knew.
West siyeed

Blogger Damelon Brinn November 10, 2019 9:41 AM  

@27, Yes, but it's infected most of the popular Linux distros now. I think Slackware might still be keeping it out, but that's not really a suitable distro for the average consumer. Some people even want to port it or something like it to the BSDs. It's like a virus.

By the way, I don't want to put down anyone's idea of forking. I'd just fork a BSD (probably FreeBSD, as the most general one) rather than Linux, for a couple reasons. One is the license. If you fork Linux, you can't ever get away from the GPL unless you literally rewrite all the code. Also, the FreeBSD source is generally considered cleaner and easier to understand, which would be critical for forking. On C programming forums, when people ask for code samples to learn from, the FreeBSD source for basic utilities like cat and head is often recommended.

I've heard some good things about Dragonfly BSD, which is a fork of FreeBSD. I really should throw that on a spare box and check it out.

Blogger James Dixon November 10, 2019 9:41 AM  

> The thing to do here is to create lots of pictures of Linux Luminaries wearing MAGA hats and flood the internet!

Dozens of pictures of Linus doing so might be enough.

Blogger OneWingedShark November 10, 2019 9:49 AM  

Rabid Ratel wrote:Time to fork the linux kernel. I have been getting some weird bugs lately, so it may be best to regress to a kernel before 5.0.0
No.
While I do agree that there should be work on Operating Systems, and even free/open-source, working on Linux would be like "polishing a turd", as Linux and other Unix-likes are badly designed. There are operating systems from the 60s and 70s that have features that are better than modern OSes, take a look at the Burroughs, Multics, and OpenVMS.

FrankNorman wrote:Now I'm wondering how soon all this convergence will result in crappy code in the kernel, because excluding it would have given some snowflake a feelbad.
Given how craptacular Windows 10 is (and the BS with Windows 8), I'm going to say five years.

Dave Dave wrote:We just fucking wanted to use a cool operating system.
Linux is not cool.

Rabid Ratel wrote:@Shimson, there are plenty of techies right here on this blog. Why wait? Let us make a Dark Lord OS that blows Linux out of the water.
I know a few people who have shown some interest in making an OS within the Ada community, including myself, perhaps I could start reaching out to them… if you're interested pop me a message (this handle at gmail). — I could also dig out my own notes and start compiling a design document for this OS; there's also some interesting things that we could do based on incorporating certain ISO standards at a fundamental level.

Blogger Rakshasa November 10, 2019 9:51 AM  

@25 do not mistake the kernel for the system utilities.

Systemd is a cancer on linux, yet something like it had to be implemented. We left the implementation to be done by the sycophants of big blobs and we not reap what we sowed.

If you don't build it they won't come.

Blogger ABroadAndAmpleRoad November 10, 2019 9:54 AM  

If you're to spend time and trouble on replacing the necessity of the Linux Foundation, put your effort into Urbit. It's not quite into userspace yet but it's getting closer and could use substantial dev attention in userspace now.

Blogger Ominous Cowherd November 10, 2019 10:02 AM  

OneWingedShark wrote:There are operating systems from the 60s and 70s that have features that are better than modern OSes, take a look at the Burroughs, Multics, and OpenVMS.

Shark, ``good enough RIGHT NOW'' always beats ``gonna be better later.'' Multics and the rest of them never got traction, and that's why. Develop one or more of those - maybe when Linux becomes totally unusable it'll get some traction, finally. Maybe not. Meanwhile, the BSDs are here RIGHT NOW.

Damelon Brinn wrote:By the way, I don't want to put down anyone's idea of forking. I'd just fork a BSD (probably FreeBSD, as the most general one)

What about NetBSD? It's been ported to everything digital, right? Is it not ready for the desktop?

James Dixon wrote:Slackware and Devuan don't use systemd.

Never tried either. I've gotten really fond of apt-get and dpkg. It would be nice to have those.

Blogger Ominous Cowherd November 10, 2019 10:06 AM  

Rakshasa wrote:Systemd is a cancer on linux, yet something like it had to be implemented. We left the implementation to be done by the sycophants of big blobs and we not reap what we sowed.

Why, please? All I remember from back then was nobody seemed to want it, we got it, and nobody liked it. I never spent much time reading the LKML, obviously.

Blogger ThatWouldBeTelling November 10, 2019 10:09 AM  

@25 Damelon Brinn:

"@19 [Shimson], I would be more inclined to invest my development or anti-convergence efforts in one of the BSDs. Linux has been making poor technical choices like systemd for a while, whether because of the growing convergence or just parallel with it, I don't know."

A study of how Pottering has been able to foist truly wretched software onto Linux would be interesting. For systemd, a lot of it is political, Red Hat having other employees trying critical software to it, and the political coup of getting Debian to use it. It was very much in Red Hat's interest, by turning the Linux userland into a Windows style hard to manage when things go wrong mess. And this was the reason I was moving towards OpenBSD before Linus was forced to drink the social justice Kool-Aid.

"The main advantage Linux has is a large number of developers to write all the hardware drivers..."

This is not an unalloyed good, for Linus refuses to create a stable kernel device driver API. So every time that changes a bunch of drivers have to be updated, and critical ones regularly rot, I've had common and uncommon standard peripherals work, then not, then work again. If you have kernel skills, which the Ubuntu Linux distribution shows is not required, you might be better off contributing to a BSD, and there's lots of cross pollination between them.

Getting back to Shimshon's original point, probably the largest amount of work in an operating system *distribution* is maintaining as many desired user level programs as possible. Few users pick a distribution because it has "the coolest" kernel, all of these are warmed over 1960s technology, they pick what can run the applications they need. Windows 10 is not cool, but just might be the best choice if the degradation of Linux accelerates. Avoid Apple, their software development is a complete and getting much worse shitshow.

Between social justice convergence, systemd, and the projected Bluewashing of Red Hat now that it's been bought by IBM, Linux is increasingly a bad bet. Damelon is sticking to FreeBSD from inertia, and because it successfully repelled a very serious SJW attack, albeit sustaining damage, I wouldn't rule it out. OpenBSD appeals to me because of its better governance as long as Theo can keep it up, and I'm for the most part willing to accept its compromises, which for many current Linux desktop users are worse than FreeBSD's.

Pick either, add device drivers if that's a skill you have or can cultivate, add software packages if you can do that. As I understand it, perhaps the best targeted work in userland is accommodating it being tied more and more to systemd, so if you have a high pain tolerance in writing shims go for it.

NetBSD has lost its mojo from what I've heard, but that could be wrong. It's unique selling proposition is that it'll run on *anything*, which is probably not our goal, which can be limited to running on x86, ARM, and perhaps someday RISC-V hardware. DragonflyBSD has its own opinions about how to do important kernel things, but is mostly an ongoing experiment in creating a distributed filesystem, which is probably not a goal of ours. If you want to stick to Linux and fight systemd, perhaps contribute to Devuan, although I note that its base Debian has one of the worst development models (details on request).

At the moment I'm using the latest LTS release of Ubuntu, which is annoying but runs some critical 3rd party software for which it's the primary Linux world target. The useland stuff I'm currently doing I make very sure runs on OpenBSD.

Blogger Crew November 10, 2019 10:15 AM  

I know a few people who have shown some interest in making an OS within the Ada community, including myself, perhaps I could start reaching out to them…

Ha ha ha ha. Well played by the community whose name is based on a lie!

Blogger Johnny November 10, 2019 10:20 AM  

This whole thing seems to me like a secular version of what happened in the Middle Ages when the Catholic version of Christianity became very corrupt and they pushed all sorts of extreme stuff as doctrine.

Blogger James Dixon November 10, 2019 10:21 AM  

> Why, please?

Because it solved problem for Red Hat corporate customers, and being an arcane mess, Red Hat is the sole source for solutions to the problems it causes.

Corporate lock in is still a thing, even with Linux.

Blogger Rabid Ratel November 10, 2019 10:22 AM  

@OneWingedShark - I suspect systemd is responsible for most of my current problems, so yes, it will have to go. The problem is that lots of software now require systemd. I have tried running without it, but it keeps getting imported by some of the packages I use. The pox on systemd!

Your suggestion of Ada is interesting, because I have looked at some secure OS developments a long time ago, and security needs to be baked into the kernel itself.

Something to look into is micro- and nano-kernels. There has been some work done a few years ago on switching between user and protected space, which is the big issue with all micro-kernels.

@Damelon Brin - I will be testing OpenBSD on my PC during the week. I will let you know how that goes.

@FrankNorman not all work is kernel related. There would be plenty of other help needed. Beta-testers would always be welcome.

Blogger ThatWouldBeTelling November 10, 2019 10:27 AM  

@36 OneWingedShark:

"Rabid Ratel wrote:
@Shimson, there are plenty of techies right here on this blog. Why wait? Let us make a Dark Lord OS that blows Linux out of the water.

I know a few people who have shown some interest in making an OS within the Ada community, including myself, perhaps I could start reaching out to them… if you're interested pop me a message (this handle at gmail). — I could also dig out my own notes and start compiling a design document for this OS; there's also some interesting things that we could do based on incorporating certain ISO standards at a fundamental level."

Isn't that an all in bet on the single company that provides a FOSS version of Ada? With what I've read is a subpar and limited runtime library?

I'd rather build something on top of seL4 (https://en.wikipedia.org/wiki/L4_microkernel_family#High_assurance:_seL4) with a Lisp, and probably a ML language family on top, with a minimum of additional C glue (seL4 is a mathematically verified C microkernel). The capability based operating systems should be investigated, although mostly because of why they failed.

"There are operating systems from the 60s and 70s that have features that are better than modern OSes, take a look at the Burroughs, Multics, and OpenVMS."

Very true, and UNIX is as much of Multics as its developers could fit on very small minicomputers. And there's the Lisp Machine, Oberon a decade later, see also what Xerox did in the 1970s. For me, a compelling anti-pattern to avoid is any project that makes serious use of C++ ("not even wrong" for group projects), which removes a lot of current work in this area.

But bringing this back to the real world, such an OS would still have to support POSIX and X, there's just too many necessary userland programs such an effort would need to support to get buy-in from a large enough base of users, absent a great deal of both money and engineering management talent. Which would still have problems getting users.

Blogger Rakshasa November 10, 2019 10:29 AM  

@40 Even without laptop issues, everything was moving towards a world where more complex logic was required for system configuration.

Systemd 'solved' that issue, and it was the only game in town.

If you are not willing to admit that there was a problem that systemd solved then you will never be able to imagine a replacement for that pile of crap.

Blogger Halibetlector November 10, 2019 10:35 AM  

This tweet about the affair gives me some hope that the entire software industry isn't doomed.

https://twitter.com/unclebobmartin/status/1192437033177931777

4300 votes. 86% say no to the question "Should someone who once posted a selfie while wearing a MAGA hat or an Obama pin be banned from attending software conferences?"

Blogger ThatWouldBeTelling November 10, 2019 10:39 AM  

@39 Ominous Cowherd:

"OneWingedShark wrote:
There are operating systems from the 60s and 70s that have features that are better than modern OSes, take a look at the Burroughs, Multics, and OpenVMS.

Shark, ``good enough RIGHT NOW'' always beats ``gonna be better later.'' Multics and the rest of them never got traction, and that's why. Develop one or more of those - maybe when Linux becomes totally unusable it'll get some traction, finally. Maybe not. Meanwhile, the BSDs are here RIGHT NOW."

Multics got a lot of traction, but internal Honeywell incompetence and politics deliberately killed it off (lots more details on request). VMS in part followed the same path, companies burned by DEC abruptly abandoning the PDP-10 line avoided it, and were proven wise when the VAX was replaced by the Alpha. There's a whole lot of good systems in the past that didn't survive because of non-technical reasons, depending on how you score falling behind because a company couldn't afford the Nonrecurring Engineering (NRE) costs to keep their system's speed competitive with Intel running something like they did, perhaps in emulation. And just bad management, Sun died in large part because their sales system would not sell to startups. When some of those companies became big, Sun was long irrelevant to them.

"[Good question on NetBSD's viability.]"

"James Dixon wrote:
Slackware and Devuan don't use systemd.

Never tried either. I've gotten really fond of apt-get and dpkg. It would be nice to have those."

So have I, and Devuan is Debian minus systemd.

Blogger Rabid Ratel November 10, 2019 10:41 AM  

@Karhu - I like the way you are thinking. Which ML language are you thinking of? I know OcaML can do some really low-level stuff, but it is also a one company pony. I will be digging into some old notes and links myself to see what work has been done to mitigate the protected userland switching, which was a nightmare on the first microkernels.

Agree with POSIX and X support.
Will shoot you an email later. need to do some digging now.

Blogger ThatWouldBeTelling November 10, 2019 10:45 AM  

@48 Halibetlector

"This tweet about the affair gives me some hope that the entire software industry isn't doomed.

https://twitter.com/unclebobmartin/status/1192437033177931777

4300 votes. 86% say no to the question "Should someone who once posted a selfie while wearing a MAGA hat or an Obama pin be banned from attending software conferences?"

This Uncle Bob character has raised the highest profile objection to the behavior of the Linux Foundation, but that and the votes he's garnered mean nothing to the amenable authorities who control so many organizations today like the Linux Foundation. They've got too much power, are not shy about using it, and will continue to cancel anyone they can.

Historically, this sort of purity spiral is only stopped by harsh violence in response to the violence it inevitably engages in, or if you believe Jim of Jim's blog, there's a few cases where the purity spiralers simply ran out of people to kill.

Blogger Crew November 10, 2019 11:02 AM  

"We will prevent you from running with scissors" ignores the fact that there are some occasions when it would be useful to run with scissors and an intelligent person would know how to recognize them.

I once came across a program written in Go that crashed in the runtime because of a null pointer dereference. Oops. I guess it's a scary world out there.

Blogger ThatWouldBeTelling November 10, 2019 11:05 AM  

@50 Rabid Ratel:

"@Karhu - I like the way you are thinking. Which ML language are you thinking of? I know OcaML can do some really low-level stuff, but it is also a one company pony."

As far as I can tell pretty much every ML is a one institution pony. Check out CAKEML, from some of the same people who brought us seL4, a verified partial implementation of Standard ML. Also look at anything Magnus Myreen has worked on, he figured out a critical meeting in the middle trick for automated verification after the Coq/CompCert approach failed on the seL4 project. It looks at the compiler generated machine code, describes what it does, and sees if that matches the higher level. If you're really concerned about security, and someday after too many people get killed by a software bug we may change our attitudes about it, look at stuff by a guy who goes by the label of nickpsecurity, for example https://news.ycombinator.com/user?id=nickpsecurity and https://lobste.rs/u/nickpsecurity

"I will be digging into some old notes and links myself to see what work has been done to mitigate the protected userland switching, which was a nightmare on the first microkernels."

As I understand it, the second generation that started with L3 and is now L4 solved that problem, has among other things to do with realizing that asynchronous requests are generally a Very Bad Idea. This seems to be very tricky, I've read that the once promising Barrelfish project failed to provide sufficient performance, and there's no external evidence of any work on it for a little more than a year, which really broke their release cadence. No listed papers after 2016, and as an academic project that's what they really care about.

Blogger Ominous Cowherd November 10, 2019 11:06 AM  

James Dixon wrote:Because it solved problem for Red Hat corporate customers, and being an arcane mess, Red Hat is the sole source for solutions to the problems it causes.

Yes, I know why systemd is a hot mess. I'm wondering what problem it solved. Obviously there was a real problem, I just forgot or never knew what it was.

Blogger Rabid Ratel November 10, 2019 11:14 AM  

@53 Karhu

It is a pity some of these projects (like SEL4) didn't get much support. I remember looking at L4 shortly after it came out, and realizing that it would be easier to maintain, because so much has been pushed out to userland that is not really needed in the kernel.

Blogger ThatWouldBeTelling November 10, 2019 11:28 AM  

@55 Rabid Ratel:

"@53 Karhu

It is a pity some of these projects (like SEL4) didn't get much support. I remember looking at L4 shortly after it came out, and realizing that it would be easier to maintain, because so much has been pushed out to userland that is not really needed in the kernel."

It's getting a great deal of support on its own merits, like for embedded aerospace applications. As the base of general purpose operating system, there's one project out there doing it for a number of L4 microkernels, but since they decided to do their own code in C++, and licence it as restrictively as they think they can get away with.... See https://genode.org/

Blogger James Dixon November 10, 2019 12:11 PM  

> I'm wondering what problem it solved.

Systems management of server farms. It provides a single and uniform management structure for managing the entire system which can scale across the entire business. What, you thought it was aimed at the end user?

Blogger ThatWouldBeTelling November 10, 2019 12:20 PM  

@57 James Dixon

"> I'm wondering what problem [systemd] solved.

Systems management of server farms. It provides a single and uniform management structure for managing the entire system which can scale across the entire business."

Can you be more specific about this? Because this is the first time I've heard that it's a multi-system thing, as opposed to a different way to manage things on a single server, the alternatives also sharing the property of "a ... uniform management structure for managing the entire system".

Not a "single one", though, you're claiming that it's (poor) reimplementations of a bunch of userland systems like DNS and time synchronization are an improvement worth its very high costs?

Blogger pdwalker November 10, 2019 12:21 PM  

@Ominious Cowherd Devuan is a fork of Debian pre systemd so you can keep your apt

Blogger pdwalker November 10, 2019 12:22 PM  

Slackware is “too hard”. Devuan is based on pre systemd debian and so is a good place to start

Blogger James Dixon November 10, 2019 1:14 PM  

> Because this is the first time I've heard that it's a multi-system thing, as opposed to a different way to manage things on a single server

Not quite what I meant, though I'll get to that in a minute.

Without systemd, every subsystem (as with your examples, dns and ntp) all have to be managed independently. And different distros implement that management differently. The system management people have to know all of them.

With systemd, as long as the distro uses it, it's a single control system with (in theory at least) a single management style which can be used across all of your systems.

And if you don't think that means Red Hat (now IBM) is working on a multi-system control interface, you haven't been paying attention. Something like Ansible that's Red Hat specific with a nice GUI.

All of this so companies get by with less qualified folks managing their systems, of course, though no one will admit that.

> Not a "single one", though, you're claiming that it's (poor) reimplementations of a bunch of userland systems like DNS and time synchronization are an improvement worth its very high costs?

No, I don't claim anything of the kind. I consider it much worse (there's a reason I use Slackware). But I'm sure that's what Red Hat claims.

Blogger Gettimothy November 10, 2019 1:28 PM  

@james dixon fellow slackware user. Much more fun than systemd.

Blogger God Emperor Memes November 10, 2019 1:39 PM  

Looks like Kim Crayton is just another minority grifter. What a piece of garbage.

Blogger OneWingedShark November 10, 2019 1:42 PM  

FrankNorman wrote:Now I'm wondering how soon all this convergence will result in crappy code in the kernel, because excluding it would have given some snowflake a feelbad.
Given how craptacular Windows 10 is (and the BS with Windows 8), I'm going to say five years.

By that "five years" I mean becomes unusable.
Over the past year/year-and-a-half my work computer (Windows 10) has crashed probably about the same number of times my 7-year old laptop (Win 7) has… and that's not counting the number of times that it bogs down so as to become unresponsive.

Ominous Cowherd wrote:OneWingedShark wrote:There are operating systems from the 60s and 70s that have features that are better than modern OSes, take a look at the Burroughs, Multics, and OpenVMS.
Shark, ``good enough RIGHT NOW'' always beats ``gonna be better later.'' Multics and the rest of them never got traction, and that's why. Develop one or more of those - maybe when Linux becomes totally unusable it'll get some traction, finally. Maybe not. Meanwhile, the BSDs are here RIGHT NOW.

Multics, VMS, and Burroughs were never marketed as personal computers either, so saying "they never got traction" is misleading when comparing them to the desktop market… also VMS did get traction and was fairly popular in commercial/governmental data-centers.

Crew wrote:I know a few people who have shown some interest in making an OS within the Ada community, including myself, perhaps I could start reaching out to them…

Ha ha ha ha. Well played by the community whose name is based on a lie!

Ada is the name of a programming language, thus it's not a lie to call the community of Ada developers "the Ada community" — regardless of Countess Lovelace's contributions to the field of CS — just like saying "the PHP community" when referring to that programming language's users.

Blogger OneWingedShark November 10, 2019 1:45 PM  

Rabid Ratel wrote:Your suggestion of Ada is interesting, because I have looked at some secure OS developments a long time ago, and security needs to be baked into the kernel itself.

Something to look into is micro- and nano-kernels. There has been some work done a few years ago on switching between user and protected space, which is the big issue with all micro-kernels.

You're absolutely right that security has to be designed into the project itself. An interesting possibility is using the SPARK subset/provers to show that the security-model is adhered to. / As to micro- and nano-kernels you might be interested in MUEN, which uses SPARK to prove its security properties.

Karhu wrote:@36 OneWingedShark:
"Rabid Ratel wrote:
@Shimson, there are plenty of techies right here on this blog. Why wait? Let us make a Dark Lord OS that blows Linux out of the water.

I know a few people who have shown some interest in making an OS within the Ada community, including myself, perhaps I could start reaching out to them… if you're interested pop me a message (this handle at gmail). — I could also dig out my own notes and start compiling a design document for this OS; there's also some interesting things that we could do based on incorporating certain ISO standards at a fundamental level."

Isn't that an all in bet on the single company that provides a FOSS version of Ada? With what I've read is a subpar and limited runtime library?

The issue is a bit more complex and not about the quality of the runtime as much as licensing, the open-source product you're thinking of is GNAT. There's three "distros" FSF GNAT done by the Free Software Foundation which has GPL with runtime and generic exceptions, and two done by AdaCore the 'Community' version which has a non-excepted viral GPL license on the runtime, and the 'Pro' version which does not.

Also, I'm working on an implementation licensed under MIT.

I'd rather build something on top of seL4 (https://en.wikipedia.org/wiki/L4_microkernel_family#High_assurance:_seL4) with a Lisp, and probably a ML language family on top, with a minimum of additional C glue (seL4 is a mathematically verified C microkernel). The capability based operating systems should be investigated, although mostly because of why they failed.
Take a look at MUEN.

"There are operating systems from the 60s and 70s that have features that are better than modern OSes, take a look at the Burroughs, Multics, and OpenVMS."

Very true, and UNIX is as much of Multics as its developers could fit on very small minicomputers. And there's the Lisp Machine, Oberon a decade later, see also what Xerox did in the 1970s. For me, a compelling anti-pattern to avoid is any project that makes serious use of C++ ("not even wrong" for group projects), which removes a lot of current work in this area.

But bringing this back to the real world, such an OS would still have to support POSIX and X, there's just too many necessary userland programs such an effort would need to support to get buy-in from a large enough base of users, absent a great deal of both money and engineering management talent. Which would still have problems getting users.

I would suggest that ignoring POSIX and X would be a good thing to look into: those services could likely be better sandboxed in some sort of VM, w/o having to shit over all the system design. (Full POSIX, for instance, requires certain directory structures.)

Blogger Gettimothy November 10, 2019 1:48 PM  

@karhu slackbuilds.org

Slackpkg is new bsd ish PKG system

Blogger The Depolrable Podunk Ken Ramsey November 10, 2019 1:56 PM  

Wearing a red hat (of all things, a Red Hat!) out in front Trump Towers now lands you in violation of software engineering professionalism according to the Linux Foundation. And of course if you wore a pink hat and blocked traffic outside of it well, you go grrl!

What to do? Well I've seen at least three Benevolent Dictators for Life get subverted or compromised in the past years. Guido van Rossum, Linus Torvalds, Richard Stallman. There's little reason to expect that Theo over at BSD won't suffer the same fate, especially if a significant jump to BSD occurs and it gains sway. SJWs are always chasing the "sway", always infiltrating locii of power to infect.

Fighting and eliminating this damned scourge of SJWs has to be the best option. Developing a counter OS is an avenue fraught with peril, remember QNX, BeOS, Hurd, etc? Adopting something like BSD is attractive, but see above, isn't it a retreat to a new front to fight the battle another day? In that case, it's a very sensible move and I'm all for it, but you have to understand it's merely tactical. We still have to prepared for the oncoming SJW assault from the new battlements.

Blogger OneWingedShark November 10, 2019 1:58 PM  

Karhu wrote:Multics got a lot of traction, but internal Honeywell incompetence and politics deliberately killed it off (lots more details on request).
I would be quite interested in hearing the story there.

James Dixon wrote:> I'm wondering what problem it solved.

Systems management of server farms. It provides a single and uniform management structure for managing the entire system which can scale across the entire business. What, you thought it was aimed at the end user?

Er… how is that different from NIS?
This is literally how my organization maintains consistency of its HOSTS files (and other config files) here.

Blogger plishman November 10, 2019 2:21 PM  

Maybe a national Linux distro maintained by a nation which attempts to have normative Christian values would be a good starting point - eg Russia's Astra Linux (https://infogalactic.com/info/Astra_Linux)

Blogger Doktor Jeep November 10, 2019 2:22 PM  

Dark OS
I'm in.
It's best feature should be to simply be bulletproof and reliable. The two things we are going to be seeing lost on Linux.
Where do we start?

Blogger Doktor Jeep November 10, 2019 2:26 PM  

Holy cow. I remember when Airport tweeted that.

Blogger Nate73 November 10, 2019 2:28 PM  

Github banned a project called C-plus which was making fun of feminism awhile back, I think they have a Code of Conduct as well.

Blogger Doktor Jeep November 10, 2019 2:31 PM  

So a one-tyrant controlled OS with Vox as tyrant is possible?

Blogger The Depolrable Podunk Ken Ramsey November 10, 2019 2:32 PM  

Ego is player in what brought us systemd. Ego has been a big driver in the Linux community since the beginning. For the first half of Linux' existence the community identified developing a "modern" desktop as its primary urgent need. The thinking was, "We need this so we can knock off Windows! This is how to win users!" This led us to things like the horrible monstrosity that is the GNOME desktop, and things like Nautilus, the $12 million file browser.

But it all wound up not mattering because where Linux really caught on came from another vector entirely. Running headless in server farms often with no desktop at all! It was all rather silly. Linux already had the XWindows system with lots of capability, much of which got subsumed in the effort to out-Microsoft Microsoft on the desktop.

There came a point where people looked around and when it came to system services (daemons) both the Macs and the Windows machines had slicker toys. Coming from UNIX, Linux had the inet.d arrangement. It was simple and well-known, but kind of clunky compared to SvcHost on Windows and launchd on the Macs when it came to launching, monitoring, stopping and re-launching system services. And with inet.d everything was mixed in with booting and root privileges, and that ball of wax.

That wasn't going to be allowed to stand. Nobody, especially not Windows, was going to get away with having slicker things than Linux. So they took Apple's launchd and started extending it, becoming systemd.

Many of the installed user base thought, "Here we go again. It's just like with the desktop. They are fixing things that aren't broken. We were getting along somehow with inet.d, now this!"

Oh, and btw, they're coming for XWindows now, too, it seems.

A large segment of Linux community has always been focused with determination on the wrong problems.

Blogger Crew November 10, 2019 2:36 PM  

I would suggest that ignoring POSIX and X would be a good thing to look into

POSIX is crazy. Unfortunately, there is a shit-load of software written that expects that interface and people don't like junking all their sunk costs.

Win32 is just as crazy. Unfortunately, there is another shit-load of sofware written that expects that interface and people don't like junking all their sunk costs.

Blogger Crew November 10, 2019 2:39 PM  

That wasn't going to be allowed to stand. Nobody, especially not Windows, was going to get away with having slicker things than Linux. So they took Apple's launchd and started extending it, becoming systemd.

Many of the installed user base thought, "Here we go again. It's just like with the desktop. They are fixing things that aren't broken. We were getting along somehow with inet.d, now this!"


I think you are a little confused. systemd (spit) is a replacement for System V Init run levels. AFAIK, inetd and inetd.conf was for starting networking services when a TCP connection came in.

Blogger Crew November 10, 2019 2:40 PM  

@70: Could we call is DarkyOS and then the Africans would be happy as well.

Blogger Crew November 10, 2019 2:42 PM  

Er… how is that different from NIS?
This is literally how my organization maintains consistency of its HOSTS files (and other config files) here.


Oh, you poor fool. Don't you know you are supposed to use LDAP in the current year?

Blogger The Depolrable Podunk Ken Ramsey November 10, 2019 2:43 PM  

Crew wrote:I think you are a little confused. systemd (spit) is a replacement for System V Init run levels. AFAIK, inetd and inetd.conf was for starting networking services when a TCP connection came in.

Here's a handy backgrounder on it:

The Tragedy of Systemd

Blogger OneWingedShark November 10, 2019 3:02 PM  

Doktor Jeep wrote:It's best feature should be to simply be bulletproof and reliable. The two things we are going to be seeing lost on Linux.
Where do we start?

Design-space, I should think.
There should also be given serious consideration to the tools used.
As the old saying goes: "Well begun is half done."

Crew wrote:Er… how is that different from NIS?

This is literally how my organization maintains consistency of its HOSTS files (and other config files) here.

Oh, you poor fool. Don't you know you are supposed to use LDAP in the current year?

I'm employed managing/maintaining equipment that goes as far back as a Motorola 68k running VxWorks. There's also a set of Solaris machines [both x86 and SPARC], a set of Windows 7 machines, and a few Windows Server 2008 machines… about the only thing I'm genuinely surprised wasn't here was a VMS cluster [though there is/was at a sister-site].

Crew wrote:I would suggest that ignoring POSIX and X would be a good thing to look into

POSIX is crazy. Unfortunately, there is a shit-load of software written that expects that interface and people don't like junking all their sunk costs.

Win32 is just as crazy. Unfortunately, there is another shit-load of sofware written that expects that interface and people don't like junking all their sunk costs.

As far as Win32 goes, it's been remarkably stable — my only real gripe against it is that MS didn't make a Win64 API and instead smashed everything together… IMO, that's where half the problem/mess comes from, the other half being that it's rather poorly organized.

But addressing that second point: there was a system called Delphi, produced by the company Borland, which had a very good object-oriented hierarchy of components encapsulating the Win32 API [and providing components/objects, obviously] called the Visual Component Library (VCL) — Taking that as inspiration for the GUI-portion of the OS would probably be really good. / Looking back further into time, Borland also had a system called TVision which was kinda-sorta similar, but did text-based interfaces… it occurs to me that it should be possible to make a component-based library which offers components that operate in the GUI and in text-mode.

Blogger Crew November 10, 2019 3:07 PM  

@79: Something in the UNIX tradition would have used make(1) and a Makefile!

Blogger Crew November 10, 2019 3:13 PM  

I'm employed managing/maintaining equipment that goes as far back as a Motorola 68k running VxWorks. There's also a set of Solaris machines [both x86 and SPARC], a set of Windows 7 machines, and a few Windows Server 2008 machines… about the only thing I'm genuinely surprised wasn't here was a VMS cluster [though there is/was at a sister-site].

Fortunately I no longer have to work with vxWorks (and the serious TCP/IP stack issue it has) or VMS. I only have to work with XXX and YYY and ZZZ.

Blogger The Pitchfork Rebel November 10, 2019 3:17 PM  

Three years ago:

https://techcrunch.com/2016/11/16/microsoft-joins-the-linux-foundation/


Embrace, Extend. Extinguish

Last year:

https://itsfoss.com/linux-code-of-conduct/

https://www.usatoday.com/story/money/2018/10/28/ibm-red-hat-acquisition/1801268002/

There's a reason Linus has been so quiet lately, of course I did see an interview where he appeared to be a man of the left.



Blogger Crew November 10, 2019 3:17 PM  

@79: The greatest concept from the video is:

A pathologically diverse range of [people]

Benno Rice would like repudiate that thought.

Blogger Doktor Jeep November 10, 2019 3:20 PM  

Did Torvalds ever take a ride with Epstein?

Blogger Crew November 10, 2019 3:22 PM  

@85: A good question.

Blogger Doug Cranmer November 10, 2019 3:22 PM  

"Ada is the name of a programming language ..."

You know that he meant smart boy.

Blogger ThatWouldBeTelling November 10, 2019 3:24 PM  

@68. OneWingedShark

"Karhu wrote:
Multics got a lot of traction, but internal Honeywell incompetence and politics deliberately killed it off (lots more details on request).

I would be quite interested in hearing the story there."

We need to recognize a heuristic for mergers where the acquired company's assets don't get destroyed. Particularly bad are when the executives from a failing company demonstrate why that happened by taking over the successful one, as happened to Netscape and Boeing. Honeywell and Multics were a somewhat different story, where Honeywell's computer line, General Electrics', from which came Multics, and SDS/Xerox were all merged together, with a bunch of incompetent managers and 3.5 fiefdoms, the 0.5 being Multics.

I don't know enough of the story to do it justice, just know a few critical tells: the first custom Multics processor development project was killed off, and the official blame was placed on microcoding the CPUs, years after IBM had taken the business world by storm with the System/360, only possible because of microcoding. So Honeywell was never competitive in CPU power for the money.

The rest was that it was a twice removed red-headed stepchild, a competitor for the bought from GE mainframes, and then the Honeywell and SDS/Xerox lines. No matter how much evidence was presented from insiders and outsiders that Multics was the *only* market advantage the company had in computers, it was ignored, and the line was deliberately killed even while Honeywell Bull in France loved it and was selling Multics systems like hotcakes. Many customers like DoD which used it to develop their budgets were unamused.

Blogger Crew November 10, 2019 3:37 PM  

@88: and the Honeywell 316 then turned up in a series of machines by another company that had five different instruction sets.

Blogger NewTunesForOldLogos November 10, 2019 4:00 PM  

No, the demands on Vox’s time are already well above 24 hours a day. We’ve reached the point where we need additional tyrants.

Blogger Shimshon November 10, 2019 4:01 PM  

Given the direction of Python, perhaps we will yet see a Perl renaissance? No one cares enough to even try to converge it. And if there were a resurgence, MPAI and wouldn't be able to use it.

Blogger VD November 10, 2019 4:19 PM  

So a one-tyrant controlled OS with Vox as tyrant is possible?

Sounds more like a job for Markku. I can be on the board.

Blogger Crew November 10, 2019 4:22 PM  

I think it should be called SaviOS!

Blogger Crew November 10, 2019 4:51 PM  

@96: How different the world would be if IBM had chosen a Motorola chip.

https://best-sci-fi-books.com/23-best-alternate-history-books/

Blogger mrpinks November 10, 2019 5:32 PM  

Debian 10 (Buster) has been good but requires many drivers to be compiled as they are missing from it, Proxmox is a great hypervisor with awesome turnkey containers. MX Linux is a good and popular distro. Ubuntu 18 is the best cloud platform (cloud init), while Ubuntu has most common device drivers compiled in the kernel, and Raspbian on the Pi. Kali for pen testing and network sniffing.
I guess the biggest problem for me lately is the requirement to find and compile device drivers to run networking devices (MX Linux and Debian 10). I dont like the look and feel of the Ubuntu 18.04 workstation. Too many things installed on setup.

Blogger VD November 10, 2019 6:15 PM  

You're banned, Dirk Manly. Stop trying to comment here.

Blogger weka November 10, 2019 6:19 PM  

Please. I want to be a vile user.

Blogger Damelon Brinn November 10, 2019 6:26 PM  

Yes, I know why systemd is a hot mess. I'm wondering what problem it solved.

On a Unix system prior to systemd, programs were started at boot by some sort of text files, usually simple shell scripts. On Linux distros they were usually all dumped into /etc/rc.d* somewhere. On FreeBSD the core OS startup files are in /etc/rc.d and the ones for everything else go in /usr/local/etc/rc.d. The init program runs these one by one to start everything else up.

So what if a package called 'before' needs to run before one called 'after'? Alphabetical order will get them backwards. If you have a Unix system administrator, he does something simple and sensible like naming the startup files '01before' and '02after' so they'll get ordered right.

But if you want end users to be able to install and run this OS and any programs that might be released for it without knowing anything or having an admin, all this has to happen automatically for thousands of different packages without a human ever thinking about how any given two might need to play nicely together. You need a system in which everything put into it is forced to play nice with everything else, and that means complexity and rules. It can't be as simple as a text file an admin can edit; you need tools to operate on it all at a distance to keep it all in sync.

Once you have this complex system in place, it becomes a hammer looking for nails. Developers start extending it to do more than program startup, making everything it touches even more complex and inter-dependent. Then Skynet takes over and kills us all.

Blogger Crew November 10, 2019 6:36 PM  

You need a system in which everything put into it is forced to play nice with everything else, and that means complexity and rules.

In essence, systemd is a Code of Conduct, and we all know what that means!

Blogger ThatWouldBeTelling November 10, 2019 6:39 PM  

@99 Halibetlector:

"Dirk Manly wrote:
Meanwhile, the x86_64 line continues only due to the huge amount of x86 executables on disk drives

It's also very very fast. PowerPC wasn't able to catch up and ARM and RISC-V still aren't capable of anywhere near the theoretical limits of an x86_64 system. I have no idea why, though."

I think POWER can do faster single threads and has a few advantages over x86_64. No ARM solution has attracted serious attention for servers, despite quite a few efforts. Probably a hard to surmount 10 year lead for x86_64 compared to the first ARM 64 bit architecture, and ARM's bread and butter is 32 bit CPUs, they're not focused on competing with Intel and AMD.

RISC-V is just too young, as best as I can tell the Privileged ISA Specification, necessary for server scale designs, was only finalized this summer. In terms of high end run Linux etc. microarchitectures, those are extremely young. I think SiFive is in the lead, certainly in shipping a development card, which has 4 64 bit single issue CPUs with 32 KB L1 instruction and data caches, but only one lower level shared 2MB cache. The next version goes superscalar with the same cache architecture.

Out of order, they've announced less than a month ago but not put on their website as a selectable option another design. They're aiming for ARM Cortex-A72 performance at 7 nm, are reported to be working with TSMC, so that would be equivalent with Intel's 10 nm node that doesn't work. Still a lot of catching up to do.

Blogger Doktor Jeep November 10, 2019 7:05 PM  

So the "Dark OS" might, ironically, have it's own unwritten CoC.
The first line saying something like "no gammas".
Sorry. Could not resist.
Wait. That's gamma behavior.
*Self destructs*

Blogger OneWingedShark November 10, 2019 7:26 PM  

Damelon Brinn wrote:So what if a package called 'before' needs to run before one called 'after'? Alphabetical order will get them backwards. If you have a Unix system administrator, he does something simple and sensible like naming the startup files '01before' and '02after' so they'll get ordered right.

But if you want end users to be able to install and run this OS and any programs that might be released for it without knowing anything or having an admin, all this has to happen automatically for thousands of different packages without a human ever thinking about how any given two might need to play nicely together. You need a system in which everything put into it is forced to play nice with everything else, and that means complexity and rules. It can't be as simple as a text file an admin can edit; you need tools to operate on it all at a distance to keep it all in sync.

Once you have this complex system in place, it becomes a hammer looking for nails. Developers start extending it to do more than program startup, making everything it touches even more complex and inter-dependent. Then Skynet takes over and kills us all.

An ordered list for execution is a complex system?
A dependency-graph for ordering that list is a complex system?

Nothing about those two issues justifies systemd.
Heck, Ada's with/limited-with dependency management is a solved problem and handles semi-cyclic dependencies. (Those where one package depends on an interface of one that depends on its own interface; by limiting the "with" to only the publicly-visible you can break the cycle.)

Blogger ZaijiaN November 10, 2019 9:48 PM  

Karhu wrote:@48This Uncle Bob character has raised the highest profile objection to the behavior of the Linux Foundation, but that and the votes he's garnered mean nothing to the amenable authorities who control so many organizations today like the Linux Foundation. They've got too much power, are not shy about using it, and will continue to cancel anyone they can.

P.S. That Uncle Bob is Bob Martin of Agile Manifesto and SOLID principles. His word still carries weight with serious coders. Someone should send him a copy of SJWAL.

Blogger James Dixon November 10, 2019 10:40 PM  

> Er… how is that different from NIS?

As I understand it, systemd is designed to control the entire system, both configuration and running status. It's a lot more than NIS. Mind you, it does so poorly (IMO) and in a decidedly non-Unix fashion, but that's what it's designed to do.

Blogger AaMcavoy November 10, 2019 10:42 PM  

Do you mind if I ask what you'd recommend to young developers looking to get into Python? Not worth it? Choose another language?

Blogger The Depolrable Podunk Ken Ramsey November 10, 2019 11:08 PM  

AaMcavoy wrote:Do you mind if I ask what you'd recommend to young developers looking to get into Python? Not worth it? Choose another language?

Python is strong choice to begin with. It runs on everything, and there's a Python inroad into every major area of computing. There's tons of books and learning material for it, too.

You can learn many principles of software engineering using Python, and the rewards come quickly. Python code is easy to write, and you can stand up useful programs very quickly.

However, Python is a scripting language. People take useful libraries and systems written in natively compiled languages like C/C++ and make wrappers exposing their functionality to Python. A good developer will know how to dig down into this native development process, too. So there's that, but still a good place to get started is with Python.

Blogger DourCdn November 11, 2019 12:27 AM  

Linux is just a evolutionary dead end. Its like those sailing ships of the 19th century, they were the pinnacle of sailing tech, sleak hulls and more and more sails with complicated rigging. But the real solution to speed was technological discontinuity, sailess engine powered ships

Blogger DourCdn November 11, 2019 12:32 AM  

There could be other directions,oberon, haiku, smalltalk. But industry doesnt want diverse tech, they want fungible commodities and developers go where there is a market demand.

Blogger Rabid Ratel November 11, 2019 1:00 AM  

Nate73 wrote:Github banned a project called C-plus which was making fun of feminism awhile back, I think they have a Code of Conduct as well.

The Dark OS need Dark Git, running on a Dark server, in the SDL's basement. All cooling fans to be installed in the eye-sockets of broken skulls the SDL has thrown out.

To move this project forward we need to discuss scope and strategies. Who can contribute and in what capacity? Will Markku be able to chair it (per the SDL) and who else of the VFM will be on board?

Blogger Sillon November 11, 2019 3:25 AM  

I advise everybody here to watch this.

https://www.youtube.com/watch?v=_-v2bI13NvA

Blogger Shimshon November 11, 2019 3:40 AM  

Fellow techies, the problem is not this or that distro, but the kernel itself. Making a new distro on top of a converged kernel doesn't change the fact that the kernel is still converged. An unconverged distro is also useful and necessary, but not really pertinent. I use Arch Linux, which is better than most in that regard, but it too migrated to SystemD years ago.

Blogger Sillon November 11, 2019 4:36 AM  

I know it is not the popular opinion, systemd has been a great improvement, it might not be perfect and it might be too ambitious in scope.

No one has come with anything better, and if you're so inclined to take it out of your system you can.

Linus is not stupid, understand he's got the enemy at home, Literally.

The kernel is too valuable now to everybody to screw it, what it should be watched out is attempts from single entities claiming to "love linux, thats why we use windows" to control it.

Blogger Sillon November 11, 2019 4:41 AM  

@98

Guess what, those magical beings, "the sysadmins", those with grey beards, who spend weeks writing scripts and file long, long bug reports.

Gasp, shock...

Like systemd because it makes system administration more effective.

Most of those magical beings have to dealt with hundreds of computers on a weekly basis.

Systemd is not perfect, I invite you to come with a better solution.

Blogger Gregory the Tall November 11, 2019 4:50 AM  

Conversion therapy for sodomites banned, conversion therapy for Linus perfectly okay.

Blogger Shimshon November 11, 2019 5:07 AM  

Re SystemD, I am with those who accept it for solving a problem well-enough, warts and all, and no acceptable alternative has been proffered.

Blogger Rabid Ratel November 11, 2019 5:15 AM  

Doktor Jeep wrote:So the "Dark OS" might, ironically, have it's own unwritten CoC.



The CoC is the Benedict CoC.

Shimshon wrote:Fellow techies, the problem is not this or that distro, but the kernel itself.

I agree, but while us techies can replace the kernel and run without systemd just for laughs, this is the start of a long term project to provide a decent system for normies to use. Security is another part of the whole system that need to be looked at.

We have to start somewhere, so while I will be testing some microkernels this week, someone needs to start thinking of packaging and branding.

How much of the Gnu base on which most distros build, is converged at this point?
There are much more to be decided on.

Blogger ThatWouldBeTelling November 11, 2019 7:20 AM  

@107 DourCdn:

"Linux is just a evolutionary dead end. Its like those sailing ships of the 19th century, they were the pinnacle of sailing tech, sleak hulls and more and more sails with complicated rigging. But the real solution to speed was technological discontinuity, sailess engine powered ships"

That may well be true, but there's a general rule of thumb that applies to the above example that's so far lacking for operating systems. The replacement has to be at least ten times better, *and that didn't happen for speed with ships*. But the total package of control when the wind wasn't cooperating, support of increased weight to accommodate iron and steel armor and better guns, and superior speed eventually became more than ten times better.

We can probably develop clean slate operating systems that are ten times better at security, but customers just don't care very much about that. I expect that to change when a software bug kills thousands of people, steam boilers went through a free for all phase with lots of explosions before getting regulated, but we aren't there yet.

Companies also don't prioritize how fast or correct software is developed. With "cloud" approaches, they don't care much about the stability of any one node, and that fine for us software people because hardware fails at a high enough rate, unless you're willing to pay lots of money for an IBM mainframe, or in times past Tandem or Stratus computers which used roughly the same techniques while adding vendor lock-in.

So what avenues exist or might exist to give a ten times improvement? The social justice convergence of Linux is a possibility, but there can be a "great deal of ruin in a nation", and GitHub is a standout example of how a company suffering a great deal of convergence can still do a good and much better job than a poorly managed one like GitLab. And the BSDs are ready to pick up the slack if Linux falters enough.

Blogger ThatWouldBeTelling November 11, 2019 8:28 AM  

@116 Rabid Ratel

"Doktor Jeep wrote:
So the "Dark OS" might, ironically, have it's own unwritten CoC.

The CoC is the Benedict CoC."

Absolutely. See the ultimately sad story of SQLite's CoC for why Rabid Ratel picked it, anyone who's allergic to it automatically disqualifies themselves.

"Shimshon wrote:
Fellow techies, the problem is not this or that distro, but the kernel itself.

"I agree, but while us techies can replace the kernel and run without systemd just for laughs, this is the start of a long term project to provide a decent system for normies to use. Security is another part of the whole system that need to be looked at."

The default assumption for some time is to be capability based, but as noted ultimately none of those operating systems made. I have a lot of artifacts stored about one line of those OSes that was pretty successful in their days.

"We have to start somewhere, so while I will be testing some microkernels this week, someone needs to start thinking of packaging and branding."

Is there any reason to extend the initial choices, prior to satisfying ourselves that they're suitable, beyond seL4 and MUEN, and only the latter if a good enough case can be made about the Ada ecosystem?

There is only so much we can bite off and chew, and while I'm not at all adverse to going the ADA and SPARK route on their technical merits, MUEN is an all in bet on the FSF's GNU GNAT. And the FSF just purged RMS. seL4 is (verified) C (through a manual translation step from Haskell) and therefore has quasi-infinitely less ecosystem risk.

"How much of the Gnu base on which most distros build, is converged at this point?"

*Linux* distros, certainly, and anything from the FSF is suspect until we see what happens post-RMS, the GPLv4 could for example be quite evil. But OpenBSD for example refuses to include anything GPLed in its base distribution, and moved from GCC to LLVM/Clang after the former was relicensed to the GPLv3. Still, there are some superior GNU or GPLed tools that are pretty much required, whether at arms length like OpenBSD or integral. I'd include Gnu Emacs, Tar, and Make in the list as tools I use every day. And I assume git for source code control.

Blogger OneWingedShark November 11, 2019 8:55 AM  

AaMcavoy wrote:Do you mind if I ask what you'd recommend to young developers looking to get into Python? Not worth it? Choose another language?
I would recommend, for all programmers, learning: Ada, Forth, and Lisp.
Ada — Gives a strong grounding in software engineering, types, and things a compiler can do automatically. (e.g. the reason there's no lint for Ada is because virtually all of lint's functionality is subsumed by the compiler and required by the standard.) Ada will teach you the value of thinking about ranges, limits, and reasoning from properties.
Forth — Gives a VERY good grounding in low-level programming, but due to the definition of 'words' (Forth analog of subprograms) as "a list of words to execute or a chunk of machine code to execute" there's the ability to rapidly abstract things. (I heartily recommend looking up a video titled something like "Forth Over the Shoulder" by Sam Falvo as it is essentially a real-time walkthrough demonstrating this.)
Lisp — It's kind of the opposite of Forth, in some ways [extremely high-level/symbolic], but exhibits the same rapid-abstraction properties.

All three of these will change how you think about programming, and together I think would provide a rich set of knowledge to draw on for the programmer of any language.

DourCdn wrote:Linux is just a evolutionary dead end. Its like those sailing ships of the 19th century, they were the pinnacle of sailing tech, sleak hulls and more and more sails with complicated rigging. But the real solution to speed was technological discontinuity, sailess engine powered ships
I think you're correct here.
The way I would like to go is with formal methods and provable code as the foundational layers of the OS. This should vastly increase both stability and security properties as well as providing high-assurance.

DourCdn wrote:There could be other directions,oberon, haiku, smalltalk. But industry doesnt want diverse tech, they want fungible commodities and developers go where there is a market demand.
This is why, I think, they are trying to push everything to some sort of C-like language, kind of trying to make everything a 'dialect' — by making all languages into C, now all programmers know all languages because "all programmers know C."

I think a lot of gain could be had by abandoning C and other C-like languages for the kernel — it simply isn't needed for low-level machine-interface: remember the Burroughs was written in Algol, and Lisp-Machines in Lisp.

Blogger Avalanche November 11, 2019 8:56 AM  

@6 "unauthorized.tech - Home of the unconverged Linux kernel.
Coming soon?"

Shimshon, are you asking someone ELSE to pick it up? Or do you think our Dark Lord has even ten seconds left in his life to add yet another huge project that "we" should be starting! Who's a retired, or financially cushioned, Linux/Unix programmer who can become the subordinate Dark Marshall of Linux?

Just as with comics -- when an avenue of escape appeared, how many prisoners of convergence ran for it? Does anyone think the unhappy but silenced Linux folks won't race for the exits if they are provided? Hell, maybe even Linus would come over -- if he is NOT being blackmailed by the 'psych' he went to ... is it protective coloring or was he broken?

"We just wanted to play our effing video games!"

Blogger Avalanche November 11, 2019 8:59 AM  

@14 "Slave mentality people who get upset by the term 'slave' used in a technical context."

And EVERY SINGLE PIECE of this hysteria is based on cultural appropriation! Why do we let these hacks and liars ruin OUR culture?

Blogger Damelon Brinn November 11, 2019 9:01 AM  

@133, Here's a nickel, kid. Get yourself a better computer.

Blogger Avalanche November 11, 2019 9:12 AM  

@25 "Linux was trying too hard to compete with Windows by being all things for all people rather than a solid Unix OS."

As I am currently trying to escape the MS concentration camp to avoid Win10 torture, I am someone who does need some version of Linux that provides a path and resources! It's not competing with Windows, it's making it so people who are finally fed-up with MS can make the leap to Linux! I studied Unix some 3 decades ago, briefly -- I CAN, I hope, make a successful leap -- except my company accting program won't go there.

I'm RUNNING a MFG company -- I don't have TIME to learn Unix and Linux Mint; and Libre Office, which I can use; and GIMP, which I do use; and some new accting program and and and.... I'm praying my new machine, if I wipe the MFG-installed Win10 off with prejudice, will let me run my accting program, under Win7Pro, in a VM.

"According to data from analytics vendor Net Applications, approximately 490 million personal computers were still running the 2009 operating system at the end of September."
That's JUST Win7; not 8.1!

"If only" there was a Linux expert/group making a version of Linux that would run Win7 and 8, and allow the programs we want and need to RUN... I'd pay a LOT! So would my main machinist whose CAD/CAM program and $70,000 MACHINE works under Win7 and NOT Win10! Now what?!

Who will be OUR King Arthur?

Blogger Avalanche November 11, 2019 9:39 AM  

@85 "Did Torvalds ever take a ride with Epstein?"

That was my first thought when Linus was remanded to Soviet re-education therapy and came out loving Big Sister!

Blogger ThatWouldBeTelling November 11, 2019 10:19 AM  

@119 OneWingedShark

"remember the Burroughs was written in Algol, and Lisp-Machines in Lisp."

And Multics in PL/1. But the low level for Lisp Machines was microcode, they didn't naively run Lisp. The closest was compiling to a byte code with the interpreter for that again being in microcode. A closer analogy is probably Xerox Parc's workstations, which were severely constrained by being 16 bit word addressed, requiring bank switching for more than 128KiB. Lisp Machines started out as full 32 bit CPUs with 24 bits of word addressed address space, 64 MiB.

Blogger ThatWouldBeTelling November 11, 2019 10:24 AM  

@123 Avalanche: For running Windows binaries on Linux, macOS, and at least one of the BSDs, check out Wine, which includes a paid option that supports more programs: https://www.winehq.org/

I use the free version every day for an essential Windows 95 program, and the occasional need that pops up in for example embedded programming. Important note if you're running Debian or a derivative of it: as of a decade ago they sliced it up into many packages resulting in baffling errors, you were strongly advised to get it in one piece from Wine's repository.

Blogger OneWingedShark November 11, 2019 3:28 PM  

Karhu wrote:"We have to start somewhere, so while I will be testing some microkernels this week, someone needs to start thinking of packaging and branding."

Is there any reason to extend the initial choices, prior to satisfying ourselves that they're suitable, beyond seL4 and MUEN, and only the latter if a good enough case can be made about the Ada ecosystem?

I hesitate to say the Ada ecosystem is better, (a) because there's a lot of things that we could do better, but which is mostly a manpower issue; and (b) because 'tooling' is the answer I often received when I polled programmers about their opinions on Ada. — I also think Ada's tooling should be better, and there's a lot of potential that is untapped: for example the GNAT system uses a GPR file for project mgt/cfg, which is vaguely Ada-like in syntax but "stringly"-based, which I think could be better done via a leveraging of generics w/ a particular structure in addition to a method of supplying values to the parameters.

And worse, too. I mentioned upthread that there simply isn't a lint for Ada due to most of that functionality being subsumed and enforced by the compiler; just because the ecosystem doesn't have tool X doesn't mean that the ecosystem is weak in that area: it may mean that that too is unneeded.

There is only so much we can bite off and chew, and while I'm not at all adverse to going the Ada and SPARK route on their technical merits, MUEN is an all in bet on the FSF's GNU GNAT.
All that is needed to escape dependency on GNAT, whether FSF or AdaCore, is a standards-conforming compiler, assuming the codebase isn't dependent on implementation-defined features.

I've compiled non-trivial Ada from 30 years ago developed on a different compiler and architecture with a modern GNAT — the only thing I had to change was (a) instances of a particular identifier, as it became a keyword in later standards; and (b) splitting one file into two because of an implementation-limitation where a file cannot contain multiple compilation units.

Avalanche wrote:I'm RUNNING a MFG company — […] I'm praying my new machine, if I wipe the MFG-installed Win10 off with prejudice, will let me run my accting program, under Win7Pro, in a VM.

[…]

"If only" there was a Linux expert/group making a version of Linux that would run Win7 and 8, and allow the programs we want and need to RUN… I'd pay a LOT! So would my main machinist whose CAD/CAM program and $70,000 MACHINE works under Win7 and NOT Win10! Now what?!

I feel your pain; in my case the work computer has had more crashes in the past year than the past 7 of my old Win7 laptop.

Karhu wrote:@119 OneWingedShark

"remember the Burroughs was written in Algol, and Lisp-Machines in Lisp."

And Multics in PL/1. But the low level for Lisp Machines was microcode, they didn't naively run Lisp. The closest was compiling to a byte code with the interpreter for that again being in microcode.

I'll defer to your knowledge here: I've only read about the internals of the Lisp-Machine and [very little] of Multics, and almost none of that was on the hardware side. Your mention of microcode does bring to memory the OpenBoot system; which, while I never got to use, sounds like it could be pretty nice.

Blogger Mercy November 11, 2019 4:22 PM  

Average Linux user here. I'd be happy to test things when needed or do some other low-skill tasks. I run Arch, using someone else's install script to give a pretty complete system. Occasionally try my hand at Void Linux. Once got OpenBSD installed, but slipped back into Linux world.

What's the target audience, if any? The more average the users, the more support they'll need to stay. Otherwise you could focus on yourselves and advanced users, then possibly make it more accessible or release a variant down the road.

Blogger ThatWouldBeTelling November 11, 2019 5:35 PM  

@127 OneWingedShark:

"Karhu wrote:

[The "tooling problem" for Ada is not much of a problem.]

There is only so much we can bite off and chew, and while I'm not at all adverse to going the Ada and SPARK route on their technical merits, MUEN is an all in bet on the FSF's GNU GNAT.

All that is needed to escape dependency on GNAT, whether FSF or AdaCore, is a standards-conforming compiler, assuming the codebase isn't dependent on implementation-defined features."

Is there any other FOSS Ada compiler implementations? If not, what I'm trying to say is that if GNAT falters, due to the converged FSF, AdaCore going out of business, whatever, we'd have to add to our workload continuing the FOSS version, or perhaps creating one based on LLVM. And I suppose we would not be alone.

This would not be an issue if we go the seL4 route and use as little C as possible to bridge from it to higher level languages like a Lisp and a ML, asserting that as soon as you can, you want to move into the non-lazy functional programming domain. Which is a very debatable point for the required POSIX and X components.

Ada would allow a lot more "safe" imperative code. And SPARK looks a lot more approachable than the mathematical proof methods used to verify seL4 and CAKEML, although those aren't exactly the same thing.

Blogger OneWingedShark November 11, 2019 7:36 PM  

Karhu wrote:Is there any other FOSS Ada compiler implementations?
Reasonably complete? AFAIK, currently no.
In-progress, my own, which as mentioned up-thread is MIT licensed.
I have heard of a few other efforts, such as HAC, but I don't know how complete they are; there are others which are stalled-out and abandoned like Dragon Egg, which aimed to be an LLVM-based Ada implementation.

Another complete (or reasonably so) open-source implementation is actually something the Ada community wants, for the reasons you mention below, plus it would be good to have non GPL-encumbered / non-convoluted situation for medium- and small-business.

If not, what I'm trying to say is that if GNAT falters, due to the converged FSF, AdaCore going out of business, whatever, we'd have to add to our workload continuing the FOSS version, or perhaps creating one based on LLVM. And I suppose we would not be alone.
Ada has one advantage that C++ (and C, IIRC) don't: the full standard is completely free and open — downloadable here and here.

This would not be an issue if we go the seL4 route and use as little C as possible to bridge from it to higher level languages like a Lisp and a ML, asserting that as soon as you can, you want to move into the non-lazy functional programming domain. Which is a very debatable point for the required POSIX and X components.
I actually have an old [almost] implementation of LISP 1.5 that I wrote as a Tutorial of Ada 2012 here, while trying to (a) teach myself the new constructs, and (b) work around a few of the bugs in the then-absolutely-new GNAT implementation.

Ada would allow a lot more "safe" imperative code. And SPARK looks a lot more approachable than the mathematical proof methods used to verify seL4 and CAKEML, although those aren't exactly the same thing.
Ada was designed to help facilitate static analysis, as the Ada 83 specification required implementation of then bleeding-edge compiler and static-analysis methods, to the point where there were some who thought it would be impossible to implement. So there's a better foundation for things like SMT provers to leverage.
Simple example: "Function X( Input : Positive ) return Natural" already encodes the fact the input must be greater-than 0 and the result is greater-than or equal to 0, these properties require no additional annotation.

Blogger ThatWouldBeTelling November 11, 2019 10:04 PM  

@130 OneWingedShark, are your current working repositories here: https://github.com/OneWingedShark?

Blogger The Depolrable Podunk Ken Ramsey November 11, 2019 11:33 PM  

Has this become a thing? Are we seriously entertaining rolling an alternate OS???

If so, well count me in.

And I say write the kernel in plain ole C. It's easier to take C and pare the language features down to the minimum of what you need -- a thin veil over the assembler. Remember, all the classic problem areas of the C standard runtime, such as found in strings.h, stdlib.h, and stdio.h, were never even needed to write a C program technically. They were always ancillary. That's why you have to include the headers even today even to do "Hello world."

You need to get down as close to the metal as you can in this sort of thing.

Blogger Rabid Ratel November 12, 2019 12:31 AM  

Mercy wrote:Average Linux user here. I'd be happy to test things when needed or do some other low-skill tasks. I run Arch, using someone else's install script to give a pretty complete system. Occasionally try my hand at Void Linux. Once got OpenBSD installed, but slipped back into Linux world.

What's the target audience, if any? The more average the users, the more support they'll need to stay. Otherwise you could focus on yourselves and advanced users, then possibly make it more accessible or release a variant down the road.


Testing, and reports on ease of use/installation are very important. There will also be a need for security testing, as this is definitely an area that needs work on most modern platforms, hence the talk of verifiable code and analysis.

Will you be able to assist in qualifying packages/programs to run on top of the OS?

Blogger Rabid Ratel November 12, 2019 3:53 AM  

The Depolrable Podunk Ken Ramsey wrote:Has this become a thing? Are we seriously entertaining rolling an alternate OS???

If so, well count me in.


I am trying to push this forward, as I am pissed off at what is happening in the OS environment, and computing in general. I need a stable, reliable OS that will not fall over its own feet just because. I am bashing both windows and Linux here.

Re your suggestion of using C, have a look at seL4. They have verified the kernel which is currently written CakeML, which is verified, and compiled into C.

We need better organizing. Best we start getting things in place before the SDL kicks us off of his blog!

Blogger OneWingedShark November 12, 2019 9:20 AM  

Karhu wrote:@130 OneWingedShark, are your current working repositories here: https://github.com/OneWingedShark?
Those are indeed my public repositories; some of them are more gist-like or experimental, or things stalled out.
The Byron project is the one that has my Ada compiler, though IIRC that version has a small bug in the Pratt-parser WRT one of the expression operators (I think I misvalued it) and may be replaced when I push the work I have on my home computer, though most of that work is in producing better documentation so that people can join the project easier.

The Depolrable Podunk Ken Ramsey wrote:Has this become a thing? Are we seriously entertaining rolling an alternate OS???

If so, well count me in.

And I say write the kernel in plain ole C. It's easier to take C and pare the language features down to the minimum of what you need -- a thin veil over the assembler. Remember, all the classic problem areas of the C standard runtime, such as found in strings.h, stdlib.h, and stdio.h, were never even needed to write a C program technically. They were always ancillary. That's why you have to include the headers even today even to do "Hello world."

You need to get down as close to the metal as you can in this sort of thing.

C will get in the way in this case.
It's now a pretty big lie that C is close to "the hardware" anymore, as this paper shows. Furthermore, a lot of the security problems we see cropping up are due directly to C's weak typesystem (which includes implicit conversions) — it's always harder to add security or correctness in than to design it with those in mind in the first place and C is notoriously hard to verify correct, which I believe to be one of the reasons that we see so many problems in OSes: aside from the orders-of-magnitude more complexity than the PCs of 30 years ago, the reliance on C for the foundational portions of the technology-stack lead to massive security-flaws.

Heartbleed, for example, would have been impossible to do on accident in Ada, requiring either Unchecked_Conversion or memory-overlays, because the natural solution is one that perfectly sizes the buffer to the message, something similar to:
Type Message(Length : Natural) is record
Header : Header_Stuff;
Data : Data_Array(1..Length);
end record;

Rabid Ratel wrote:Testing, and reports on ease of use/installation are very important. There will also be a need for security testing, as this is definitely an area that needs work on most modern platforms, hence the talk of verifiable code and analysis.

Will you be able to assist in qualifying packages/programs to run on top of the OS?

Formal methods & provability as mentioned above in sel4, MUEN, and SPARK are a good way to reduce (not eliminate) the need for testing. The paper Safe to the Last Instruction is an excellent illustration.

Blogger Scott Alfter November 12, 2019 11:48 AM  

Rabid Ratel wrote:It is a pity some of these projects (like SEL4) didn't get much support.

It appears to be too late for seL4, at least. I hadn't heard about it until today, but a few minutes' poking around revealed this:

https://docs.sel4.systems/Conduct.html

The cancer has already spread.

Blogger Scott Alfter November 12, 2019 11:50 AM  

James Dixon wrote:Without systemd, every subsystem (as with your examples, dns and ntp) all have to be managed independently. And different distros implement that management differently. The system management people have to know all of them.

With systemd, as long as the distro uses it, it's a single control system with (in theory at least) a single management style which can be used across all of your systems.


https://xkcd.com/927/

Blogger ThatWouldBeTelling November 12, 2019 12:20 PM  

@134 Rabid Ratel:

"The Depolrable Podunk Ken Ramsey wrote:
Has this become a thing? Are we seriously entertaining rolling an alternate OS???

If so, well count me in.

I am trying to push this forward, as I am pissed off at what is happening in the OS environment, and computing in general. I need a stable, reliable OS that will not fall over its own feet just because. I am bashing both windows and Linux here."

My biggest question is, "Why not instead contribute to making OpenBSD better?" Because starting with a microkernel is both fraught with danger (see Barrelfish), and a *great* deal of work to provide POSIX and X on top of, unless of course we grab code from a permissively licensed existing OS like OpenBSD. Probably tens of man years worth of work, unless someone can sketch out a clever path.

"Re your suggestion of using C, have a look at seL4. They have verified the kernel which is currently written CakeML, which is verified, and compiled into C."

CAKEML is a separate thing started up after seL4 had gotten some distance. Both are verified using Isabelle/HOL: https://en.wikipedia.org/wiki/Isabelle_%28proof_assistant%29

"We need better organizing. Best we start getting things in place before the SDL kicks us off of his blog!"

Time to set up a mailing list, and make sure it gets backed up. I can help with the latter.

Blogger ThatWouldBeTelling November 12, 2019 12:29 PM  

@136 Scott Alfter:

"Rabid Ratel wrote:
It is a pity some of these projects (like SEL4) didn't get much support.

It appears to be too late for seL4, at least. I hadn't heard about it until today, but a few minutes' poking around revealed this:

https://docs.sel4.systems/Conduct.html

The cancer has already spread."

seL4 still is based on serious math that to my memory no strong whamen have contributed to. If the proofs are correct, and Isabelle/HOL verifies an implementation down to the assembly language, it's still good, to a substantial degree it's inherently resistant to SJW corruption. Getting cast out of the official community for wrong think? Lots of ways of handling that, including using an email alias to just read the mailing lists.

Also remember the example of GitHub, and even Linux for now, there can be a "great deal of ruin in a nation."

Blogger Mercy November 12, 2019 3:49 PM  

Rabid Ratel wrote:Testing, and reports on ease of use/installation are very important. There will also be a need for security testing, as this is definitely an area that needs work on most modern platforms, hence the talk of verifiable code and analysis.

Will you be able to assist in qualifying packages/programs to run on top of the OS?


I can install packages and follow basic instruction, may need guidance the first time. Very average linux user. But willing to help however I am able. Repetitive tasks are fine.

If I'm qualified to qualify, please sign me up.

Also, I see Distrotube and many of his commenters have noticed the trend. He's a relative normie, probably conservative, quit a job after being called racist or something. Now dipping his toe in the MAGA arena. Not a fighter, but maybe an indicator of sentiment. https://www.youtube.com/watch?v=a8TcBeltvk4

Blogger The Depolrable Podunk Ken Ramsey November 12, 2019 10:56 PM  

OneWingedShark wrote:C will get in the way in this case.

It's now a pretty big lie that C is close to "the hardware" anymore, as this paper shows.


I'm not totally convinced about that. C is yet not that far removed from writing in assembly. Yes, we have many more instructions now such as SIMD and we have parallelism like we never had before. But this problem is affecting all languages. We have things like CoArray Fortran and facilities like OpenMP and OpenCL which are piled on top of C for parallelism and vectorization. So damn it if C and Fortran still aren't leading the charge at modelling processors. Yes, they do so much worse than before and far more messily, but still there they are. They are still the gorillas and ones with wide support and user bases.

I take what you have said to heart. I'm agnostic to any final selection of tools, really. You like Borland's VCL? I've got the latest Delphi installed right now. And you are 100% correct, Borland did Windows better than Microsoft (Bill Gates used to rail about this). Hell, let's roll it in Pascal. It wouldn't be the first time anybody did that, for example the original Macintosh OS.

I would say that when we are talking about the kernel space of an OS, we ARE talking about something that's completely segregated from user space (or at least it ought to be, haha!). The OS is really there to merely provide interfaces to hardware, and keep itself in the background as much as possible. The simpler and the more scalar the better. The wild west (and the handcuffs) are for the user space that emerges from it.

Blogger James Dixon November 16, 2019 9:48 AM  

> Systemd is not perfect, I invite you to come with a better solution.

There were several alternative solutions. Systemd won out because it's from Red Hat. But for desktop users, systemd was completely unnecessary. It's a big business solution.

The insistence on merging business class systems and end user systems has never worked out well for end users.

> this is the start of a long term project to provide a decent system for normies to use.

See above. It needs to be an end user system, not a business server system. The needs aren't the same.

> check out Wine, which includes a paid option that supports more programs: https://www.winehq.org/

I.e., Crossover. https://www.codeweavers.com/products

It doesn't run everything, but many business programs are supported, and "for a small fee", you might persuade them to work on something you need.

Post a Comment

Rules of the blog

<< Home

Newer Posts Older Posts