Memory-safe sudo to become the default in Ubuntu(trifectatech.org) |
Memory-safe sudo to become the default in Ubuntu(trifectatech.org) |
I found an alternative implementation that doesn't rely in being a setuid binary like systemd-run0 much more interesting from a security perspective, but I am no security expert.
I think the main benefit of eliminating setuid binaries is that you can forbid them system-wide (e.g. via mount flags), as a hardening measure.
So yes, I am not saying that privilege escalation bugs are impossible if you have a different architecture, but like Lennart argues is that it makes them much more difficult to happen, especially because creating a proper setuid is difficult. Also there is a bunch of things that makes sudo especially tricky to implement correctly.
Eventually the bug (and associated vulnerability) was patched. I think.
Ubuntu continuously updates itself without permission, killing apps and losing previous state. You have the Javascript based Gnome window manager that is always bugging out. The Ubuntu packages, drivers and kernel are laughably behind Debian and even further behind mainline. Ubuntu continues to morph into something I don't believe in.
That all said, Rust is not a smoking gun for incorrect application logic. It could still happily incorrectly execute stuff with the wrong permissions or blow something up badly. I think it's also a bad idea to offer it as a drop-in replacement when clearly features are still missing since a long time [1].
[1] https://github.com/trifectatechfoundation/sudo-rs/issues?pag...
Nitpicking, but I thought Android was the most widely deployed Linux OS around...
I don't have a problem with it, specifically. Seems odd that they don't advertise it, though.
[1]: https://www.sudo.ws/security/advisories/
[2]: https://www.sudo.ws/security/advisories/unescape_overflow/
This makes me wonder:
1) Would a hypothetical "sudo-lite" with these features removed lead to better security without a rewrite?
2) If these features are useful in the real world, will a Rust rewrite of sudo inevitably gain these features over time and end up with similar problems?
By all means, write a better, safer, smaller, more secure version of sudo, but don't have command be sudo or sudo-rs, name it something else.
Somehow "maintainability" has never been something I'd ever associate with Ubuntu. Is it a reference to their source tree? That'd make a lot more sense than if they were referring to the OS itself.
The Linux port has not been maintained for 3 years. Has unmerged rowhammer fixes and generally a yolo auth system best described as "dangerous". You are better off using a well maintained project, that includes the CVEs^Wwarts.
It's a mistake to think that `doas` on Linux is the same as `doas` on BSD.
- https://github.com/Duncaen/OpenDoas/issues/106
- https://github.com/slicer69/doas/issues/110
I have a hard time recommending doas over sudo on Linux when the issue has been fixed in sudo but not in doas.
Here is my doas config:
cat /etc/doas.conf
permit nopass jane as root
Sudo: #ALL ALL = (root) NOPASSWD: C_ZFS
I have no idea what is going here, not sure what ALL means, why root is in (), etc.
> some features of the original sudo will not be implemented in sudo-rs if they serve only highly niche use cases
That’s what 99% of distros default to, and it’s simple.
This side steps the issue which is "Does Rust help you make software more correct?" No one is arguing that Rust is perfect. There are plenty of bugs in my Rust software. The question is only -- are we better off with Rust than the alternatives?
> I think it's also a bad idea to offer it as a drop-in replacement when clearly features are still missing since a long time [1].
Your example is the Github issue page?
Look -- I agree that, say, uutils/coreutils missing locales may frustrate some users (although I almost never use them). But "close enough" is more the Unix way than we may care to realize. But especially in this instance, because sudo is not POSIX (unlike locales which are). A distro is free to choose any number of alternatives.
Ubuntu wants to lay claim to "the Rust distribution" and it's hard to blame them when Linux wants to lay claim to "the Rust kernel".
Entirely untrue. It may happen, but there is zero consensus to port Linux to rust. Not even the tiniest bit.
There is a lot of embedded knowledge in existing implementations, Rust deals with just one small class of bugs but drops a lot of this knowledge in the process.
I would generally be in favour of just introducing better memory management to C/C++ and hard enforcing it in particular repositories.
It does default to installing security updates automatically. However, this is completely configurable.
It isn't reasonable to have different default behaviour - otherwise the majority of users would be vulnerable from a security perspective.
If you want different behaviour, just configure it as you wish.
> The Ubuntu packages, drivers and kernel are laughably behind Debian and even further behind mainline.
This just isn't a reasonable description of reality.
Unless you're referring to an Ubuntu LTS, in which case, of course it is: that's the entire point of Ubuntu LTS, and Ubuntu users have the choice of using the six monthly non-LTS releases if they want more up-to-date packages, which Debian users do not have.
Man, but have you personally tried to disable it?
Did you stop apt-daily.service apt-daily.time apt-daily-upgrade.service and apt-daily.timer? Did you repeat the same but masking and disabling those services. Don't forget to repeat that for unatended-upgrades.service. Even after that whenever our CI fails an apt-get we have a pstree output to figure out what other dark pattern canonical came up with.
This whole debacle made me consider RedHat for my next install, and I use Ubuntu for almost 2 decades. It became unreliable in servers.
Dont get me started on the lack of security updates on "multiverse" packages which starts to include more and more packages and thus LTS means less and less. This is not innocent but so you buy Ubuntu One.
A better default behaviour would be to alert the user and allow them to choose to indefinitely defer by "accepting the risk". Some setups, rightfully or wrongfully, have a very long running time and cannot be restarted.
> If you want different behaviour, just configure it as you wish.
I'm not sure if it changed, but they made it extremely difficult on purpose. You can stop snap from updating, but then lots of other things also break.
> This just isn't a reasonable description of reality.
It's my experience with packages I use.
What? Is this some snap thing because apt sure as hell doesn't do this without you configuring it explicitly.
There's no control of when snaps update, Ubuntu has explicitly said they will never add this.
There was no way to disable snap auto-updates until just last year(-ish?) when Firefox finally announced they would no longer support snaps and started telling people how to tear them out and replace them with native packages or Flatpaks. Low-and-behold, Ubuntu suddenly got the feature to disable automatic snap updates. After saying explicitly they would never allow it for years, and telling high-uptime users to instead block snap daemon network access via the firewall to avoid it.
The bigger problem is upgrading packages deliberately but being surprised by the results. My team's current favorite is the upgrade process itself suddenly having new interactive prompts breaking our scripts.
I've never seen this happen and I've run Ubuntu in production for years. Apt does not auto-update unless it's configured for unattended upgrades — and both Debian and Ubuntu allow you to configure unattended upgrades in apt. And unattended upgrades via apt should not kill running user processes or cause data loss.
The Ubuntu packages, drivers, and kennel are laughably behind Debian.
This is just plain wrong — even for the steelman argument of Debian unstable or testing, which are not intended for general use. Debian unstable and testing are on kernel 6.12. Ubuntu 25.04 is on kernel 6.14.
Debian stable, meanwhile, is on 6.1. Ubuntu has the far more-recent kernel.
I don't know what you mean by "drivers" — there aren't separate drivers on Linux from the kernel; they're shipped in the kernel. Ubuntu's are also more recent than Debian, since the kernel version is more recent.
With respect to packages, obviously I can't check every package version, but e.g. coreutils in Ubuntu are on 9.5, released in March 2024; systemd on Ubuntu is a version released this year (and until last month Debian unstable and Ubuntu were identical); gcc is identical; etc. While Ubuntu occasionally lags Debian unstable, it's not by much.
If you compare to actual Debian stable, it's not even close. Debian stable is ancient.
And ultimately... Why are you using Debian unstable? It's called "unstable" for a reason. It receives basically no testing. Even the "testing" version is more stable, and that's not intended to be stable at all and doesn't necessarily receive security updates. Ubuntu is less-stable than Debian stable, but far more up-to-date; Debian testing is less-stable than Ubuntu... And usually still not even as up-to-date. Debian unstable is basically untested; if you want that you'd be better served by a rolling release distro like Arch where the packages are going to be way more up-to-date anyway.
The Debian wiki cautions against treating unstable or testing releases as general purpose, so I truly don't think even this steelman is viable. [1] In fact, they refuse to even call Debian unstable a "release" since there are no release practices associated with it and the code is effectively untested.
Ubuntu is nowhere near my favorite Linux distro, but claiming it's more out of date than Debian is just FUD.
Debian is very very stable — at least, Debian stable is — and people love it for that. But the tradeoff is that everything in it is ancient. If you want something that's like Debian, but more up-to-date but slightly less stable — that's Ubuntu. If you want a rolling release, that's Arch. (And of course, there are even more-different distros like NixOS or ostree-based ones; there's the Red Hat universe of RHEL and the closer-to-bleeding-edge Fedora; etc etc.) Using Debian unstable is either a magnanimous act of sacrifice in order to help test future Debian versions, or it's self-harm.
Personally if I wanted to use a Debian-derivative on the desktop, though, I'd probably use System76's PopOS, which is basically a cleaned-up Ubuntu with some nice GNOME extensions. I'm more curious in the future to try out ostree-based distros, though, like the various Fedora Atomic ones, since they have nice rollbacks without the user-facing complexity of NixOS.
I tried the latest Ubuntu and it seems to be targeted at either containers or desktops. Everything I wanted to set up networking wise was a pain for my little non standard configuration.
Ended up wiping it and installing Debian instead.
As for this Rust thing, first question that comes to my mind is what features are missing from this new godly impervious to hackers by default implementation.
Ubuntu seems to find it necessary to always invent some new way of doing a standard thing. Like how they use netplan for networking, a tool they invented themselves for a task that already has industry standard options available, is missing basic features those alternatives have, and adds nothing the alternatives don't also have (including any better usability). They do this all the time, and have to eventually be dragged into the modern era when they finally get sick of having no community support for their one-off inferior tool.
In particular I'm just waiting for snaps to finally die. But at least that has some technical possibilities the alternatives don't, they just aren't functionally available in snaps yet. In another 20 years, if Ubuntu keeps at it with their unconfigurable, hardcoded, private snap registry and slow limited advancement snap portals-equivalent implementation, they might even have half as much functionality and packaged tools as Flatpak current has today.
---
If you want a decent server, Debian is a better option, even though they have some finnicky choices, and its enough like Ubuntu you might have some cross-ober familiarity. Some of the old standbys like Fedora aren't good options because of their frequent update schedule and lack of long term support, but there are also some very good niche options if you can dig a lot more.
Also worth noting: if you want to keep the server working, you should plan on pretty much everything being in containers. It adds some complexity to what you're doing, but keeps each little experiment isolated from the others and avoids polluting the global system.
One of my former colleagues used to install Ubuntu servers. I replace them with Debian when I get the chance. I was already blacklisted for Snap, so I can't re-blacklist them for going uutils and sudo-rs, and that's sad (as in Bryan Cantrell's famous talk).
Bit like how tomato is technically a fruit but everyone knows that in the context of a supermarket it’s a vegtable
What issues did you have with snap packages?
I am not aware of any problems with snap packages.
They have some pretty nice features compared to traditional packages:
https://snapcraft.io/docs/get-started#p-19156-connect-an-int...
If snap packages have caused problems for you, what were the problems?
Idk, just speculating to maybe get the thought process
There's no "Linux operating system". Linux is the name of the kernel. Android is an operating system, GNU/Linux is, things like BusyBox/Linux are. They're all operating systems that use the Linux kernel.
Linux is not a Unix https://en.wikipedia.org/wiki/Unix-like
[1] because it doesn't abstract the hardware platform, doesn't multiplex hardware, nor protect software principals from each other https://www.youtube.com/watch?v=36myc8wQhLo&t=8m48s
Agree. The question whether to rewrite and/or whether to use any new implementation should take this fact into account.
> Rust deals with just one small class of bugs but drops a lot of this knowledge in the process.
Hard disagree. Rust explicitly deals with several very important classes of bugs (memory safety and concurrency), and also aids correctness via other helpful design features like tagged unions and immutability by default. But Rust, the language, does not drop any knowledge in the process, though any decision to rewrite in any language may drop some knowledge, and/or may create new bugs, in the process.
> I would generally be in favour of just introducing better memory management to C/C++ and hard enforcing it in particular repositories.
This is really easy to say, but in practice it just hasn't worked out, and there is loads of empirical evidence to back that up.[0] It is not as if market incentives don't exist to create better C/C++ code.[1] If you have a way to do it better, I have no doubt Google, half a dozen other hyper-scalers, and the US government will pay you handsomely for your solution. But, at this point in time, if this is the solution, I'm afraid it's time to put up or shut up.
[0]: https://www.usenix.org/conference/enigma2021/presentation/ga... [1]: https://www.darpa.mil/research/programs/translating-all-c-to...
It really wasn't too difficult to get high reliability for memory management in C/C++ which is also concurrency safe (I have active projects running for years like this). The difficulty was enforcing it so that you are reduced to a subset of the language and it has to be followed with discovery at compile time.
The trap I'm concerned we are falling into is the "just re-write the C/C++ project in Rust". I still believe the solution is in enforcing better practices at compile time.
It’s not that simple.
And if we're talking about memory bugs (which we were up to now), then definitely no.
But run0 is new C code anyway so I don't see how your claim is relevant.
Even with your explanation I don't think it fits here.
There's no valid reason cyber security people would take a well known idiom and repurpose it as you imply, and a quick Google suggests they haven't done this.
... But I did not say there was a consensus to port Linux to Rust? I'm sorry you misunderstood.
Now, why would Linux want to lay claim to being 'the Rust kernel' and how is that different than Linux being rewritten in Rust? I believe that there are many reasons why Linus chose to give Rust for Linux a chance. I believe at least one of those reasons is mindshare. If Linux chose not to experiment with Rust drivers, then that mindshare might go somewhere else.
>> Ubuntu wants to lay claim to "the Rust distribution"
Notice, Ubuntu is doing a similar thing. Canonical isn't porting all of Ubuntu to Rust. It is picking and choosing bits which they can, and would like, to move to Rust. Why? Probably for similar reasons. They want to seen as friendly to the Rust mindshare.
This isn't a thing. Linux isn't laying claim to any such assertion.
If you want to know Linus's reasons, then read the LKML. He's quite open in all of his thoughts in this regard, and it has nothing to do with labeling Linux 'the Rust kernel'.
I don't know if this is some weird sort of advocacy, or you're just listening to a lot of over the top rust people, but nothing you're saying here is real.
Again, very sorry you misunderstood me. However, I am now pretty certain one of your difficulties is that you stop short of reading my next sentence, and then my next sentence after that. See my quoted comments above. I made very clear these are strictly my beliefs.
> If you want to know Linus's reasons, the read the LKML.
Perhaps when I said "I believe" I was being too subtle about what "my beliefs" are or what "beliefs" mean.
I hope you would agree -- just because one has expressed certain technical reasons/desires does not mean that there were not any unexpressed social reasons/desires, or what philosophers also call "higher order volitions" (long term volitions, or volitions about volitions), for one's actions.
Now -- I do not know but I do believe there may be social reasons for Rust being adopted in the Linux kernel, because I have read the LKML and I have reasoned about why Linux is experimenting with Rust drivers from things Linus and others have said.
Feel free to disagree, of course, but, in the future, please make more of an effort to avoid mischaracterizing me again.
The original unix process abstraction was extremely simple; the entire spec is a few pages.
The problem is that Linux keeps adding more and more levels of Rube Goldberg machine to its security model, so now literally no one understands how a default minimal install of, say, Ubuntu works.
Adding a magic daemon that runs stuff as root to this pile of complexity probably won’t help. Ripping out almost all the cruft that’s accumulated over the years, and adding back something sane (maybe BSD jails) would work a lot better.
The non-daemon has to parse just as much in addition to making itself secure. Actually it needs to parse more things in more complex ways.
I'm personally moving to Debian. It's 99% how Ubuntu used to be and most Ubuntu stuff is just a .deb that is relatively compatible.
Sure. It's just a one line change in the configuration file (/etc/apt/apt.conf.d/50unattended-upgrades). Or, if you're doing a mass deployment, just don't install the unattended-upgrades package.
> figure out what other dark pattern canonical came up with
The mechanism is inherited from Debian. It isn't Canonical's architecture.
If you want to hack internals to do things in a more complicated way, then that's up to you, but you can't then complain that it's unnecessarily complicated.
That answer shows you have not seen that pattern fail.When that fails or is overwritten by an update, remember my comment.
Ironically you just added another way to configure a simple thing, proving my point.
https://www.theregister.com/2021/02/11/microsoft_azure_ubunt...
> In practice, there are few installations that use sudo-ldap. Most installations that use LDAP as a directory service and sudo have now opted for sssd, sssd-ldap and libsss-sudo.
> The Debian sudo team recommends the use of libsss-sudo for new installations and the migration of existing installations from sudo-ldap to libsss-sudo and sssd.
That would be terrible UX and is exactly contrary to Ubuntu's philosophy, which is to do the right thing by default.
The alternative is to bombard the user with questions that they're generally not in a position to understand, and force them to receive an education on stuff that doesn't matter to most users before they can use their computer.
Even in Windows (or at least it used to be), the decision to perform an update now was a user decision. Just killing off applications without warning is the worst UX ever. Randomly killing stuff off is the opposite of what I want my OS doing.
> The alternative is to bombard the user with questions that they're generally not in a position to understand, and force them to receive an education on stuff that doesn't matter to most users before they can use their computer.
It doesn't have to be like that. It could be: "Do you want to update now? The following programs are affected and will be restarted: X, Y, Z. [Learn more]" The answers could be "Yes", "Remind me on next boot", "Remind me later" (offers common delays, i.e. 1 hour, 1 day, 1 week).
What is should never do is take the power away from a user. I saw an Ubuntu user's system restart their snap programs in the middle of delivering a conference presentation without warning. That was the worst way that could have been handled.
It's been years since they added warnings for upcoming snap updates. There's also "refresh awareness", which defers updates (to a limit, with warnings before exceeding the limit) while a user is using an app.
Here’s a simple implementation: https://github.com/TheMilkies/rut/blob/main/rut.c
(Though it doesn’t clear the environment unless I’m missing something - they should probably replace the call to execvp with one to execvpe, and add a null pointer to the end of the argument list).
The problem of setting up root’s environment and parsing the command line is left to the shell in both solutions (the thing I linked doesn’t indirect through a root login shell).
There’s also the config file, but that’s the same for both.
Similarly, the system could be running some SEL derivative or be using a capability system that causes non-standard behavior from system calls, but the daemon has the same problem.
I apologize, I do not bookmark these issues, but maybe I should start doing that? In any case, you will find logic bugs which may raise the question "is it really worth the rewrite?".
> I apologize, I do not bookmark these issues, but maybe I should start doing that?
Yes. If you want to point out problems, it really helps if you can point at specifics.
> In any case, you will find logic bugs which may raise the question "is it really worth the rewrite?".
There's a cost/benefit question, but note that the mere presence of some bugs doesn't make a rewrite worthless.
OpenBSD did this with their doas utility:
That's not to say that it won't create other ones.
This is how dpkg and apt have worked in Debian and Ubuntu pretty much since their inception. Look into debconf, dpkg and ucf configuration to learn how to integrate these with your automation. The mechanisms for this have existed for decades now and have not substantially changed in that time.
If you're installing software from Debian/Ubuntu repos, you can only use aptitude or apt to my knowledge. Other tools give you the ability to install DEB files you already have, and manage what's on your system currently. And aptitude and apt are both well known for never having had a "stable" scriptable interface. In fact they themselves tell you that their commands are not stable and should not be used for scripting, despite no alternative mode or application existing.
Recently Ubuntu moved to apt 3 as well, which massively overhauled the tool from apt 2. All those scripts people wrote to use apt 2 (because there was no alternative) broke recently when they now had to use apt 3.
> In fact they themselves tell you that their commands are not stable and should not be used for scripting, despite no alternative mode or application existing.
No, that's just the apt command, not the apt-get command, and the manpage for apt tells you exactly how to do this instead. It's clearly documented, so your "despite no alternative mode or application existing" claim is simply ignorant.
Please read the documentation and learn how to use the tooling before criticizing it and misleading others with claims that are outright wrong.
There's this weird definition of "operating system" which means "some sort of platform with a shell and little commands like cp, ls, etc.". That's just what POSIX tells people an operating system is, not the ultimate truth. It doesn't have to be that way.
Drop the "mischaracterising" routine. You're asserting specific things, so expect to get challenged when they're nonsense.
If I said "Coca Cola wants to lay claim to being the best cola soda in the world", I'm sure you would also say to me: "Patently untrue. Where exactly in Coca Cola's public statements are we to find that statement?!", instead of, perhaps reflecting, and asking yourself -- is that a reasonable belief for one to hold, given Coca Cola's marketing?
If I am not conforming to your expectations, perhaps it is because your expectations that need a reset.
You stated this as fact, as an active statement and goal. It isn't. At all. It's made up fantasy.
Trying to reframe things after, by changing that statement into "oh, that's just an idea I had!" and then blaming others, is invalid and dishonest.
You seem to want to blame others for your made up, untrue statements being challenged. Give it a rest. Your attempts to blame shift will gain no traction here.
You stated something as fact that is not. You were wrong to do so. You are wrong to blame me for pointing it out. You are wrong to continue complaining.
You are wrong.
Clear?
He was comparing to "normal" sudo, but sudo-rs have the same problems he highlighted anyway
run0 does not (and instead relies on systemd).
---
To answer your next question: setuid, while historic, is a bit weird, and is disabled in some environments, e.g. NoNewPrivileges.
If your bar is “I’ll tolerate such crap”, you may as well run your desktop session as root.
You can enforce a boundary between root and an account that never elevates though. And as far as I understand hardening sudo helps with that.
Maybe you can't realistically do it on Linux, because Linux doesn't care about desktop security and doesn't have the kind of privileged GUI that you need.
* Privileged gui to display and approve what will run as you mentioned. In Linux, non-privileged gui is used to request permission to do... something.
* Executable verification. Let's say someone with user level permissions swaps out the program you wanted to run for a malicious one. The replacement would not pass certificate checks which would be revealed in the privileged gui. Maybe you could supply an old version with known vulnerabilities though idk?
* Kinda commandline parameters verification. In Linux-land you can pass parameters to the program those could be intercepted with user-level permissions. In Windows this would be displayed in the privileged gui. However! The parameters are only visible if you click "show more" which I would guess <1% of people actually do.
As an example of the last point I tested making a shortcut to cmd.exe and used Run As Administrator. Then I changed the shortcut to C:\Windows\System32\cmd.exe /c calc and again did Run As Administrator. Opens calculator instead of command prompt with identical approval gui (except with show more).
So stop doing that!
Going back to the topic of the discussion, making sudo more resilient prevents a whole class of possible bugs (see CVE-2021-3156). You may not like sudo, but it is very commonly used to manage access, prevent accidental errors, or improve accountability on Linux servers and workstations. It is not going away. Therefore, improving sudo is a worthwhile goal and a thing to be celebrated.
>If it was provided by a daemon, built into systemd, or anything else
Yes, this is also dangerous.
So does your OS.
Ugh. Well, I suppose it must seem very unfair to live a life without the benefit of figurative language and/or subtextual meaning. Know that I'm praying for a cure.
Good luck sir or ma'am!
You've now crafted multiple lies, over your muktiple replies, to explain away your untrue statements. This too is wrong.
And it's also why it mostly has not happened for most people.
>sudo more resilient prevents a whole class of possible bugs
Good, but this doesn't fix the easiest way to escalate privileges as an attacker through sudo. Memory safety doesn't help people who run "curl | sudo bash" to install a random program from the internet.
>It is not going away.
But if work is done it could become very niche and eventually stop getting new usage over time.
I don't think you can help those people, unless you seriously lock down the machine to the level of iPhone. Neither Android nor Windows have "sudo" for example, and yet they can get malware just fine... all the difference is instead of the password, user clicks on confirm button a few times.
I advice to (1) think about how you use sudo (2) think what would you replace it with and (3) think how the replacement can be abused by malware.
Here is one example:
(1) Today I used "sudo" to run docker: "sudo -Eg docker docker run..."
(2) the sudo-less replacement would be to add myself to docker group, or switch to rootless docker
(3) If I add myself to docker group, malware can trivially escalate by starting a privileged container. If I switch to the rootless docker, I am weakening security boundary between my primary user and the docker containers I run.
In text format, all-caps or the use of an exclamation mark are often used to indicate shouting.
So, it is not unreasonable for people to read your sentence as such.
It can't be enforced on Linux because `sudo` can be trivially MitM'd, but you can't do that on Windows because it's just a click.
Being a setuid binary means that sudo also suffers from attacks where an attacker runs `sudo ./malware` and then convinces the user to authenticate
That's why the OP said that's not an enforceable security boundary. If the user is capable of attaining superuser privs, you can trick them, regardless of how attaining those privs is implemented.
I didn't interpret OP's comment like that. I think he was saying you can't enforce the boundary at all even if users don't get tricked.
That's true on Linux because the sudo UI can trivially be MitM'd by malware. You can't do that on Windows so trivially.
[0] https://cyberdom.blog/abusing-the-windows-update-stack-to-ga...
Unless you mean something like "capabilities are always inherited, and there is no way to increase them", which would mean that sudo-like scripts are impossible, and you need to start all-new session as root user for admin actions. Good news, it's already possible in linux - just don't install "sudo", and maybe set up SELinux to disallow all escalation. Turns out no one wants to do this though.
there's no god-mode who can do everything. there's no universal kernel level. micro means only minimal things happen in the kernel or at root level, like routing messages.
even windows understood that those times, when they did split Administrator from System. linux followed then with system users for certain services. but capabilities only allow certain calls at certain times. there's no setuid 0, only cap this and cap that for this admin user for a certain time range.
BTW, SElinux is the default on a proper linux like Redhat. I use it daily. I also have to adjust the ssh caps, not just the firewall settings. Everybody wants that, just not the debian nobs.
It doesn't need to be possible.
>change system configurations
You can have an settings app to configure the system.
>then you have Android
Which is much farther ahead than Ubuntu on security. Ubuntu needs to play catch up.
>but then you have certainly not enabled everything a user can do with sudo.
The goal is not to be able to do everything. The average user doesn't need to be able to do anything. Especially with their regular account.
Anybody who finds themselves using sudo is already well off the beaten path, by their own choice. There's nothing wrong with that.
* The user may never perform the action (would require sudo, but the user doesn't have sudo rights)
* The user may sometimes perform the action (i.e. only after authenticating with sudo)
* The user may always perform the action (is always implicitly authenticated)
"Being root" is just another name for the last option.
What fourth alternative do you have in mind?
No, it's not. Take for example ping. If we want users to be able to always be able to use ping does that mean they need to be root? No, it doesn't. A privileged part of the OS can handle doing the raw socket and the unpriviledged user can talk to that part of the OS.
The key point is that some operations that require privileges are okay to expose to a user, but giving the user privileges for everything is dangerous.
The beaten path is the defaults. Windows and MacOS particularly.
An agent logically has all the capabilities necessary to do what the agent should be able to do. The sum of capabilities of an agent indicates "the worst that can happen" if the agent is malicious. It makes sense that if a network service is malicious, all networking activities can be subverted. Still, the storage activities shouldn't be subverted, and of course the network service wouldn't have the storage service capability. However, if a user is malicious, anything could go wrong that the user is normally trusted to not make go wrong. Correspondingly, the user must have an expansive sum of capabilities.
Capabilities are themselves simple, but that is the mechanism perspective. Access control policy is an entirely different beast, and any mechanism at best minimizes the risks.
However, were were talking about "sudo" and this is purely interactive tool (I am sure some people run sudo non-interactively, but we can all agree it's a bad security practice). So, how does this nice capabilities vision mesh with the idea of "administrator user"?
From what I see, it's not much different. Instead of "sudo" which verifies user identity+permissions and then gives out "root" or some other user, we have some other hypothetical tool ("become-admin"?) which verifies user identity+permission and gives out powerful system-changing capabilities. Frankly, I don't see much difference - it's the same security model either way.
(A related argument I've seen is "you should not need sudo for common tasks like update software or configure network" - but on modern Linuxes, you don't need sudo for that already, this is done via dbus + polkit calls. No "proper microkernels" required, that is already deployed everywhere.)