Is Profile 17.1 a move toward a Linux monoculture?

[saellaven]

[Naib]

have a read of the gnome blog I linked, that covers it all from the GNOME side of things. Trying to dig through the gentoo ml at the time is a nightmare (and that is what GNOME found when they were tying to understand the noise from Gentoo about GNOME requires systemd).

The only reason I was following it was, while I was never a GNOME user (WindowMaker -> Openbox and now AwesomeWM)... my wife needed something more "classical" when she used my PC so I kept gnome installed for that... Ironically she got along with early gnome3 (be it painfully slow) than my openbox???? I had to yank GNOME and then rebuild soo much as GNOME had wormed it's way into a lot of what I was using in openbox (unexpectedly) when it suddenly depending on systemd... I spent a lot of time trying to understand whether it was a gentoo political move, a mistake, short-term etc... Everything pointed back to upstream so I janked the last bit of gnome off my machine and just installed xfce panel for my wife
_________________

[Fitzcarraldo]

[CasperVector]

[CasperVector]

[Anon-E-moose]

Elogind and eudev made systemd people a little more honest and in line.
When there were alternatives, the systemd/RH cabal had to change their dictatorial behavior.

Before elogind became stabilized, they would change the inner workings of systemd to make logind harder to extract as a standalone binary, the same was going on with udev before eudev got popular.

Remember LP's "this is a wakeup call, gentoo" (paraphrased as I don't remember the exact verbiage) re. udev being used with anything other than systemd.

Edit to add: Going to what Naib has said before, if systemd had been properly engineered, then all the components would have had clean boundaries, where they could either be replaced (if needed) or used standalone (without a ton of work). In the past I did plenty of work with complex systems where there were boundaries, where you wrote to the interface, for example the db part, and the db could be easily changed by using something else and writing to that same interface. It's not difficult, but it does require a little preplanning, not "it's tuesday and I've decided to add xyz to the program today, just because". Of course if systemd were to have been done that way, then people could have chosen the init part, dropped the binary logging, chose logind, dropped network, etc, with a monolithic bloc then it's take it all. So from that aspect systemd wasn't in the make linux better mode it was in the "control" linux mode. And here we are.
_________________
PRIME x570-pro, 3700x, 6.1 zen kernel
gcc 13, profile 17.0 (custom bare multilib), openrc, wayland

[Naib]

[krinn]

[Naib]

[krinn]

What i mean by that Naib is that you cannot define an engineering failure because you made a non reliable car when your boss asked you to make it like that.
Sure from external eyes, it looks a failure (everyone would assume at least reliabilty should be part of the goal in a car), but it was a goal (and in this case, the best engineer will make one with a predictable date of failure ; and the worst enginner would make a reliable one...).
I'm not sure systemd design was make like that by poor choices or randomness.

[GDH-gentoo]

[Fitzcarraldo]

[Anon-E-moose]

Thanks for the link, Fitzcarraldo, I took a glance at it, but will have to save a thorough reading for later, but does look interesting.

Having said that, I don't think there would have been any problems (certainly not the ones we have now) if systemd were only an init & init-script replacement.

It's all the other things added, trying to be the middle man between the kernel and the user that makes people wary of it.

I'm not a big fan of LP, having seen his previous work, and his lack of following through with said work to completion.
But I haven't followed systemd from the initial concept so I can't say if it was his idea he sold or not, so I'll take your word for it.
But whether he sold the idea to RH or RH came to him, it's immaterial, RH has used it to gain unprecedented control over multiple linux distros and it only looks to get worse, in the short term, rather than better.

For a comparison, one only has to look at windows to see that constantly adding things doesn't make it better, it makes it more complicated to do it right.
No clear separation between subsystems, has a tendency to cause it all to fail, in usually unpredictable ways, which I think will manifest in the not too distant future for systemd. As more and more people use it, in ways that LP hasn't foreseen, there will be more BSOD equivalents. But that's all just my conjecture.
_________________
PRIME x570-pro, 3700x, 6.1 zen kernel
gcc 13, profile 17.0 (custom bare multilib), openrc, wayland

[depontius]

[Fitzcarraldo]

[Anon-E-moose]

I remember when windows 3.1 and earlier was just a windowing desktop on top of dos, it could be swapped out and there were alternatives to it, using dos as the base. But MS didn't like that, it didn't get them lock-in so they hid dos deep within windows so that it was hard to separate it. There were vendors who sold various add on utilities, and over time MS added them either with EEE or buying them out or creating their own version that wasn't as good as the alternatives, but was free. And we all know where that ended up. We don't need that for linux to be "successful", we don't need an RH only world. I prefer having choices, whether an init, an editor, a windowing desktop, whatever. Monocultures hate choices, "how dare those pesky end users want something else other than what we offer"

Edit to add:
And I'm not saying RH is wrong to do the things they do, they're a business, they look at things from that aspect.
Nor am I saying LP/systemd people are wrong with building what they want to build, everyone has the right to scratch an itch.
_________________
PRIME x570-pro, 3700x, 6.1 zen kernel
gcc 13, profile 17.0 (custom bare multilib), openrc, wayland