|
|
Log in / Subscribe / Register

Use cases

Use cases

Posted Aug 9, 2024 20:44 UTC (Fri) by bluca (subscriber, #118303)
In reply to: Use cases by gidoca
Parent article: Distinguishing Debian testing from unstable

The use cases and purpose of os-release is documented at https://www.freedesktop.org/software/systemd/man/latest/o... and explained in the blog post that was published when it was introduced https://0pointer.de/blog/projects/os-release


to post comments

Use cases

Posted Aug 10, 2024 13:20 UTC (Sat) by ju3Ceemi (subscriber, #102464) [Link] (16 responses)

Ha ! All of this crap, to finally ends with systemd.. I did not saw that coming. You should have said it earlier, it would have made things more understandable.

Use cases

Posted Aug 12, 2024 8:44 UTC (Mon) by LtWorf (subscriber, #124958) [Link]

Well he's a systemd contributor https://github.com/orgs/systemd/people

Use cases

Posted Aug 12, 2024 10:59 UTC (Mon) by zdzichu (subscriber, #17118) [Link] (14 responses)

One of the first links in the article points to systemd's documentation.

Is your hostility due to systemd, or becase Luca is a Microsoft employee?

Use cases

Posted Aug 12, 2024 11:27 UTC (Mon) by Wol (subscriber, #4433) [Link] (4 responses)

It's probably because Luca is extremely opinionated. Nothing wrong with that, but when his opinion clashes with someone else, we get fireworks.

I've tried to keep out of this, but my opinion here is that SID is in a permanently broken state, so what's different about os-release being broken?

Also, os-release is external to Debian, and it appears to be the considered opinion of Debian that - inasmuch as it applies to SID - the definition of os-release is broken.

(I would also comment that - having been heavily involved with LSB in the early days, imho the entire premise of LSB was badly broken - as in its approach provided the opposite of what was actually required. os-release seems to be a good example! What LSB *should* have provided is a way for the 3rd party to ask the system "do you provide X", for various values of X, and in way that could be passed in to the package manager so that X could be installed. Put otherwise, "here is a list of my dependencies, please package manager would you install them".)

os-release is meant to tell 3rd-party packages what it can expect the base system to provide for it. As far as SID is concerned, that is "a broken linux system that isn't even guaranteed to boot". In which case, what's the point of even having os-release, if the system isn't guaranteed even to get as far as a state where said 3rd-party package can read it!

Cheers,
Wol

Use cases

Posted Aug 12, 2024 12:05 UTC (Mon) by mb (subscriber, #50428) [Link]

The problem is that both opinions are valid.

Luca's view is valid.
Sid can be used as if it were its own rolling release.
In fact, I have been using it for the last two decades as such.
And no, it does not break all the time. Only occasionally. :)

But it is also valid to see Sid as just a precursor place for the packages that will eventually end up in testing without a modification.

The difference in thinking is the upgrade path thinking (Luca) versus the package migration path thinking (mine; and apparently also Debian's).
These paths are orthogonal.

If Debian decides to mark os-release in Sid with the testing codename, that is perfectly valid in the package migration path world view.

The person who wants to change this established policy is responsible for providing good real world examples showing why the change improves things.

So far I didn't see anything that would convince me. Pointing to the presence of a sid workaround in a random shell script does not prove that sid is wrong. The shell script can also be wrong to use the codename in the first place.

It is needed to explain to people why the script can't do it in another way instead of pointing to random github searches that carry not a single bit of information to resolve the issue. (just to mention one ridiculous example).

Use cases

Posted Aug 12, 2024 13:24 UTC (Mon) by bluca (subscriber, #118303) [Link] (2 responses)

> It's probably because Luca is extremely opinionated. Nothing wrong with that, but when his opinion clashes with someone else, we get fireworks.
> I've tried to keep out of this, but my opinion here is that SID is in a permanently broken state, so what's different about os-release being broken?

The fireworks happen not because of opinions clashing, but because of opinions and facts being conflated. You are obviously entitled to hold an opinion such as "sid is bad, I will never use it". You are not entitled to your own facts such as "sid is permanently broken" - because it's factually not, as there are not only users, but entire downstream distributions using exclusively sid, to whom the fact that it is permanently broken will sound as a surprising piece of news. Not only that, but the entire development cycle of Debian heavily relies upon there being sid users, doing QA.
Likewise, one is entitled to say "os-release is bad", or "I have no use for os-release", or "Debian should not use os-release", or "nobody should use os-release". Those are legitimate opinions that I disagree with. "There are no use cases for os-release" is not an opinion though, it's just a false statement, given even a cursory search finds literally hundreds of thousands of examples of usages.

Use cases

Posted Aug 12, 2024 14:44 UTC (Mon) by Wol (subscriber, #4433) [Link] (1 responses)

> You are not entitled to your own facts such as "sid is permanently broken" - because it's factually not, as there are not only users, but entire downstream distributions using exclusively sid, to whom the fact that it is permanently broken will sound as a surprising piece of news. Not only that, but the entire development cycle of Debian heavily relies upon there being sid users, doing QA.

PLEASE DO NOT MISQUOTE ME!

In fact, it's pretty obvious you did a knee-jerk and (I admit I often do it) read what you THOUGHT was there, not what ACTUALLY WAS there. Just read the phrase before where I wrote "sid is permanently broken" for a perfect example!!!

And given that the *whole point* of SID is to upload stuff and see what breaks, I strongly suspect that - at any AND ALL times - there is a significant minority of people for whom SID really IS broken. Maybe for 99% of people, 99% of the time, they never notice, but SID tells you you should *expect* to be left holding a handful of pieces every now and then!

Cheers,
Wol

Use cases

Posted Aug 12, 2024 15:11 UTC (Mon) by bluca (subscriber, #118303) [Link]

I mean, if multiple people understood the same thing, maybe the phrase was just confusing? Of course you are allowed to "think" that sid is permanently broken, but the fact that it isn't still stands. Yes it's unstable, and yes things go wrong, but that's the same everywhere, on most rolling distros.

Use cases

Posted Aug 12, 2024 11:33 UTC (Mon) by ju3Ceemi (subscriber, #102464) [Link] (8 responses)

Neither
I think the /etc/os-release file is fine
What I still wanted to know is the use cases behind the proposal: what would brings the change, from a user perspective

After reading so many walls of texts, it ends with "because systemd", which was .. unexpected. Maybe there are not serious reasons ? That would explain the "we don't care" from TC.

Use cases

Posted Aug 12, 2024 13:14 UTC (Mon) by bluca (subscriber, #118303) [Link] (7 responses)

There have been literally tons of text exchanged here and elsewhere, and yet not a single time "because systemd" was said by anybody.

Use cases

Posted Aug 12, 2024 13:22 UTC (Mon) by ju3Ceemi (subscriber, #102464) [Link] (4 responses)

"So what are use cases for this that have been brought up?"
And you answered : "The use cases and purpose of os-release is documented at https://www.freedesktop.org/software/systemd/man/latest/o... and explained in the blog post that was published when it was introduced https://0pointer.de/blog/projects/os-release"

Which I read, btw. And I have found no use cases for this proposal.
I found no use cases but was randomly insulted by the post author.

At this point, the whole topic must be some kind of joke or something ?

Use cases

Posted Aug 12, 2024 14:35 UTC (Mon) by bluca (subscriber, #118303) [Link] (3 responses)

> "So what are use cases for this that have been brought up?"
> And you answered : "The use cases and purpose of os-release is documented at https://www.freedesktop.org/software/systemd/man/latest/o... and explained in the blog post that was published when it was introduced https://0pointer.de/blog/projects/os-release"

Yes, those are the websites where the specification lives, so what?

> Which I read, btw. And I have found no use cases for this proposal.

If there are no uses cases, why is it used literally hundreds of thousands of times all over the place on Linux?

> I found no use cases but was randomly insulted by the post author.

Where and how were you "insulted" precisely?

> At this point, the whole topic must be some kind of joke or something ?

Ok, you are not here for a serious conversation, my bad

Let's calm down

Posted Aug 12, 2024 14:39 UTC (Mon) by corbet (editor, #1) [Link]

Can we (everybody involved in this discussion) try to calm it down a bit? The mailing-list discussion has already happened, we don't need to reproduce the whole thing here.

Use cases

Posted Aug 12, 2024 14:40 UTC (Mon) by ju3Ceemi (subscriber, #102464) [Link] (1 responses)

"If there are no uses cases, why is it used literally hundreds of thousands of times all over the place on Linux?"

The whole topic is about testing vs unstable.

Nobody is arguing against /etc/os-release.

Use cases

Posted Aug 12, 2024 15:13 UTC (Mon) by bluca (subscriber, #118303) [Link]

> The whole topic is about testing vs unstable.
> Nobody is arguing against /etc/os-release.

Well, actually this _is_ about os-release. It's pretty much what the whole thing is about...

Use cases

Posted Aug 14, 2024 4:01 UTC (Wed) by Kamilion (subscriber, #42576) [Link] (1 responses)

> and yet not a single time "because systemd" was said by anybody.

You said/implied it.

Here, let me, someone with no dog in the race, break it down for you, since you're not seeing the obvious.

> "The use cases and purpose of os-release is documented at [...snip...]/systemd/[...snip...]

Systemd...

> explained in the blog post that was published when it was introduced https://0pointer.de/blog/projects/os-release"

Lennart's blog, typically about... Systemd...

> "Oh look, a DDI ( https://uapi-group.org/specifications/specs/discoverable_disk_image/ )! I wonder what it is?"
[...snip...]
> "Fedora 40, cool. Oh look, another DDI! I wonder what this one is instead?"
> $ sudo systemd-dissect image.raw

DDI, eh? Hm, where did that come from again? ( https://0pointer.net/blog/the-wondrous-world-of-discovera... ?)

So at face value, "please help me correct the output of this systemd tool" is clearly what people are perceiving.

And don't you even start with me over "but that's just where the specifications live". Or try insinuating that I haven't interacted with mkosi. Or try insinuating that I, the maintainer of ubuntu-customizer, lack familiarity with image artifact generation, customization, cataloging, or any of the other ad-homonyms that I expect to be lobbed at me.

Regardless of any of that, my personal opinion is that sid should not declare itself as trixie.

I've already had enough annoyance this year dealing with PCI compliance audits demanding I install an OpenSSH version only available in unstable to meet their tool's policy preference.

Use cases

Posted Aug 14, 2024 9:32 UTC (Wed) by bluca (subscriber, #118303) [Link]

> You said/implied it.

Nope.

> And don't you even start with me over "but that's just where the specifications live".

It is exactly that though. It's just where the spec lives. I am sorry for any remaining haters, but it's not 2013 anymore, just because something has "systemd" somewhere in the url doesn't automatically mean it's the work of the DEVIL, it's time to move on. Even distributions that do not use systemd, and are very strongly against it like Alpine ship an os-release file:

NAME="Alpine Linux"
ID=alpine
VERSION_ID=3.20.2
PRETTY_NAME="Alpine Linux v3.20"
HOME_URL="https://alpinelinux.org/"
BUG_REPORT_URL="https://gitlab.alpinelinux.org/alpine/aports/-/issues"

Source: https://dl-cdn.alpinelinux.org/alpine/v3.20/releases/x86_64/alpine-minirootfs-3.20.2-x86_64.tar.gz

Because it's just a simple and effective standard format for identification of vendor trees, and has no requirement that are related to systemd as a project, either to use it or create it, and that's very much intentional. What more proof do you need to believe that this spec is independent of systemd?

> DDI, eh? Hm, where did that come from again?

From here: https://uapi-group.org/specifications/specs/discoverable_disk_image/


Copyright © 2026, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds