Please wait...
So add Arch to the Sh*t List?
Arch made my shit list when they symlinked /bin to /usr/bin, indicating they don't understand the use case for having /usr on a separate mount point. Unix fail.
oooo ouch. Ya i agree.
Tue Feb 04 2025 02:20:26 UTC from zelgomerSo add Arch to the Sh*t List?
Arch made my shit list when they symlinked /bin to /usr/bin, indicating they don't understand the use case for having /usr on a separate mount point. Unix fail.
( mni rant, but its Linux related at its core )
So RedHat is in process of integrating AI ( IBM's granite series ) into Gnome.
yay? Nah, no thanks. Didn't like gnome much anyway, it seemed too childish and inconsistent.
And yes, that is me saying that. While we all know i love most of AI stuff and have been messing with it for decades long before the current LLM revolution, i feel its the same with all digital stuff, it has its place, and that is not everywhere 'just because we can'. Ironically, when i was a kid, i struggled rationalizing moving to electric guitar from acoustic.. even tho i was to the point of designing CPUs ( on paper and trying to create them using EEproms ) as a hobby.. Even had a mechanical analog clock thru college and into early adult hood ( until a cat knocked it off my dresser and broke it ).
And yes, Gnome can suck my balls at this point.
Well darn, i always strive to be different than the rest :)
if i had the option, id be off-grid out in the appalachia or something.
Engineers living low tech is an actual trend.
Arch made my shit list when they symlinked /bin to /usr/bin,
indicating they don't understand the use case for having /usr on a
separate mount point. Unix fail.
The original rationale, as I remember it, was that you could have one or the other mounted read-only and/or served up as a shared volume. And that totally made sense at the time, when disk was expensive and the operating system binaries only changed every couple of years. There are a lot of reasons why that kind of thinking doesn't make sense anymore:
1. Every machine gets updates frequently, but not at the same time. It's easy to get out of sync.
2. If you're sharing, there are better ways to deduplicate. Many filesystems and storage arrays will simply do it for you.
3. It is the current year. Disk is cheap. Mounting /usr from a shared volume doesn't really make sense anymore.
That having been said, the symlink thing seems silly to me as well. Combining them doesn't really seem to accomplish anything.
So RedHat is in process of integrating AI ( IBM's granite series ) into
Gnome.
GNOME was created at the direction of Microsoft by their mole Miguel de Icaza to deliberately fragment the Linux desktop. I still can't get it to work the way I want to -- that is, with the taskbar on the bottom and no other panels. I want my computer to look like a computer, not like a phone or a Mac. Yes, I know about "dash to dock" and "dash to panel" but they're both unstable. Most recently I ran updates on my older laptop and the whole desktop crashed on startup every time after that.
Fortunately I had just heard about OpenMandriva so I wiped the machine and installed it. Haven't looked back since. Their implementation of KDE Plasma is really polished and works great. And it isn't built by people who hate me. What's not to like!
Monolithic Microsoft-think. Just like systemD
That having been said, the symlink thing seems silly to me as well. Combining them doesn't really seem to accomplish anything.
3. It is the current year. Disk is cheap. Mounting /usr from a
shared volume doesn't really make sense anymore.
This is never a good reason, for any feature or resource. I don't know why people keep repeating this.
You're basically saying, "we made the movie theater wider so now we can fit fatter people" instead of "now we can fit more people."
Believe me, I've done it. Sharing /home is an absolute win. Sharing /usr/local is a win if everyone is on the same architecture. But sharing /usr is a disaster.
3. It is the current year. Disk is cheap. Mounting /usr from a
shared volume doesn't really make sense anymore.
I really hate that argument because I usually get it when somebody wants to discard something good for whatever reason, and the only argument they can build is "It is old".
I personally like that there is a /usr so you can differentiate the things that are needed for recovery / single user mode and those who aren't.
Even so, I still contend that mounting a shared /usr is madness in the modern world where OS updates ship almost daily. Mounting a *dedicated* /usr on, for example, a diskless client system might make sense, in which case the server or storage array can handle deduplication and save disk that way.
My experience here -- and remember, I am a seasoned data center architect with decades of hands-on -- is that diskless systems have peaked and are now receding. The reason for this is that Hyperconverged is now all the rage. Once system designers realized that every node can contribute both a share of the compute pool *and* a share of the disk pool, dispensing with the cost of an expensive storage array, that became very popular very fast. This will be the case for as long as "cloud" remains popular.
But as every experienced IT bod knows, our industry is a big set of pendulums that eternally swing back and forth. So if you don't like the way something is, just wait a while.
The collapse is beginning. I wonder if CUDA works in *bsd yet....
"Christoph Hellwig, a Kernel maintainer, has revealed that Linus Torvalds privately expressed his intention to merge Rust code into the Linux kernel."
I meant to say accelerating.. pottering started the collapse.. that evil SoB.
The collapse is beginning. I wonder if CUDA works in *bsd yet....
"Christoph Hellwig, a Kernel maintainer, has revealed that Linus Torvalds privately expressed his intention to merge Rust code into the Linux kernel."
2025-02-20 22:15 from Nurb432 <nurb432@uncensored.citadel.org>
Subject: The death of linux
The collapse is beginning. I wonder if CUDA works in *bsd yet....
"Christoph Hellwig, a Kernel maintainer, has revealed that Linus
Torvalds privately expressed his intention to merge Rust code into
the Linux kernel."
I thought this already happened. I could swear last time I looked in the linux tree I saw something Rust in there. I guess it could have been some userspace lib or tool, I don't renwmber.
Who would have imagined that crap language attracts crap people.
( ok, perhaps not ALL rust people as in any group there are exceptions, but stereotypes win today )
Rust people were problematic in other ways.
Oh, and when i promote myself to world emperor, after i level the new housing and return it back to farmland, next thing will be to mandate python, everywhere. And to take Pottering out to the public square perhaps after that.
python, everywhere. And to take Pottering out to the public square
Ew, Python?? Might as well take me to that same public square.
There's a time and a place for everything, I guess. Python turns out to be a really good language for systems management (which is why Ansible is built from it). I still have a pretty firm belief that the kernel should be built from a single language. That language is C and it doesn't make sense to start mixing other languages in. However, it seems that ship has sailed and we're going to have Rust code in the kernel (starting with 6.1 so apparently it's already here).
I haven't decided whether I'm going to learn Rust yet. So far I haven't become involved in any projects that use it. The part of my job that involves writing code is all systems management stuff so it's mostly Python along with some IaC languages like Ansible and Terraform. At home I build Citadel exclusively in C (with browser-side code in JavaScript of course) so there's really no pressing need to learn another language right now.
If code is going to be distributed, one should also be concerned with the complexity of setting up a build environment for anyone wanting to compile the program themselves. I'm proud of the fact that most of what gets distributed from here requires little more than gcc and gmake. Consider the Linux kernel now: it requires not only a C compiler but also Rust and over 70 extensions to Rust just to get the NVMe driver built. Then you've got the fact that the C portions use the GNU back end and the Rust portions use the LLVM back end. That's a pretty complicated build environment. When you get to that point, people start building the tooling to set up dedicated build environments for specific projects, instead of just using "whatever compiler is already on the system."
(Side note: Linux kernel ought to switch from gcc to clang so the whole thing can be LLVM)