|
Gazpacho posted:yosproject: convince lkcl and xah lee to collaborate on and submit a kernel patch i had not heard of xah lee before. googled the name, found a hn thread, and welp i found something which fits in a number of yospos threads seansoutpost 197 days ago | link Xah,It really sounds like you need a bitcoin address. Coinbase.com (YC S12) is a great place to start. Once you have some coins, you will find no shortage of people in the bay area willing to buy them off of you. I would even be willing to run a btc fundraiser on your behalf. I don't have a lot of rep on here, but it's pretty solid in the bitcoin space. Google Jason King Bitcoin or Sean's Outpost Best of luck, man. Stay strong.
|
#
?
Nov 30, 2014 00:52
|
|
|
|
| # ? May 23, 2024 19:44 |
|
|
unironic ironic computerforum is always the best place for actual Good Effortposts on computer things on the internet
|
|
#
?
Nov 30, 2014 01:34
|
|
|
linux has one more month to make this thread come true. will it make it? taking all bets. odds start at ten trillion to zero, that's right, a bet of zero dollars pays out ten trillion if linux wins
|
|
#
?
Nov 30, 2014 01:39
|
|
|
Sagebrush posted:linux has one more month to make this thread come true. will it make it? taking all bets. odds start at ten trillion to zero, that's right, a bet of zero dollars pays out ten trillion if linux wins it will. then about that time some collection of neckbeards will claim all the neurotypicals ruined it and then fork it into linuf: linux is not user friendly.
|
|
#
?
Nov 30, 2014 04:47
|
|
|
pseudorandom name posted:shouldn't we use a multi-dimensional system with separate axes for Hans Reiser, Andre Hedrick, and Jeff Garzik? i googled andre hedrick for old times sake (so many bizarro paranoid lkml flamewars) and... welp.  didn't expect that ending. i guess it fits but holy poo poo. at least it turned out infinitely better for his family than reiser's meltdown did.i hope there's nothing sinister about jeff garzik i don't know, he seems only mildly nutty
|
|
#
?
Nov 30, 2014 06:09
|
|
|
BobHoward posted:i googled andre hedrick for old times sake (so many bizarro paranoid lkml flamewars) and... welp. BobHoward posted:i hope there's nothing sinister about jeff garzik i don't know, he seems only mildly nutty
|
|
#
?
Nov 30, 2014 06:52
|
|
|
Controversial blog post published, let's see what death threats I get today. Somewhat hastily written, and I could easily write 2x as much, but I wanted to keep it small and contained. Feedback welcome. http://blog.mecheye.net/2014/11/why-package-managers-are-not-my-ideal-software-distribution-mechanism/
|
|
#
?
Nov 30, 2014 22:49
|
|
|
Suspicious Dish posted:Controversial blog post published, let's see what death threats I get today. Somewhat hastily written, and I could easily write 2x as much, but I wanted to keep it small and contained. Feedback welcome. Aren't apps workable because the runtime is guaranteed, where this isn't true across Linux distros?
|
|
#
?
Nov 30, 2014 23:00
|
|
|
i think i am probably a member of the audience most likely to get care-mad about packaging, but your essay fails to anger me.
the only bone i really want to pick: quote:Today’s package managers do not see the OS independently from the applications that make it up: all packages are just combined to create one giant filesystem tree. this is a feature, not a bug. using a single namespace for the filesystem works really, really well in combination with rpath. don't like dealing with system libraries? no problem. set your rpath to only use libraries from /opt/endlessm/lib and /opt/endlessm/usr/lib. rpm actually supports "relocatable" packages but no one uses them. it's hard to write the scripts for them, sysadmins are not familiar with how they work, and relative paths in the rpath field is just big pain in the rear end.
|
|
#
?
Nov 30, 2014 23:02
|
|
|
The issue is with "they build one filesystem tree" is that you can't upgrade apps and your OS independently. You have to update in lockstep, or you fail to upgrade both.
|
|
#
?
Nov 30, 2014 23:05
|
|
|
Captain Foo posted:Aren't apps workable because the runtime is guaranteed, where this isn't true across Linux distros? it is and it isn't. at a base level, the kernel ABI and glibc are pretty stable. you can set an environment variable and get consistent behavior from any kernel + glibc newer than the one you originally coded against. above that, lsb compliance mostly works. lsb is what assures you that /tmp is safe to use and you will find most filesystem things where you think you will find them, and certain libraries are guaranteed to be present. if you just "package" everything as a tarball or self-installing executable that dumps poo poo into /opt, depending on no system libraries, distributing on linux isn't impossibly difficult. this is exactly what many vendors do. poo poo that isn't really easy/stable/guaranteed:
|
|
#
?
Nov 30, 2014 23:07
|
|
|
If you're curious, Endless has our own (really bad) application bundle format, which we unpack into /endless/, just to make you extra mad.
|
|
#
?
Nov 30, 2014 23:07
|
|
|
i agree with the premises but not Poettering's conclusion, especially if it makes btrfs mandatory and makes heavy use of the moral equivalent of overlay mounts (a hideous hack). In particular, invoking a root helper to mount an untrusted brtfs filesystem image on top of your current FS namespace every time you want to run a third party app is insane: btrfs has been around for what, eight years now? and it doesn't even work reliably in benign environments, never mind hostile ones. I am very reluctant to trust my data to btrfs if they can't even get it right after this many years. also let's not double-down on the Docker poo poo and force people to install every single Linux distro simultaneously because plenty of fuckheads out there are still idiotic enough to target Ubuntu (and all its gratuitously incompatible and shittily designed desktop bits and pieces) and the Firefox guys will base their poo poo on Red Hat and the Libreoffice guys will base their poo poo on Debian or whatever. This isn't progress, this is a godawful loving mess.
|
|
#
?
Nov 30, 2014 23:07
|
|
|
Suspicious Dish posted:The issue is with "they build one filesystem tree" is that you can't upgrade apps and your OS independently. You have to update in lockstep, or you fail to upgrade both. i don't see why that is necessarily the case as a third party vendor, if you target a specific OS version, you have a much easier time using system libraries. if choose not to use system libraries, you can target a lowest-common-denominator kernel and glibc version and ship your entire runtime inside. that is to say, while upgrading the apps and the OS concurrently is often a practical requirement, that's not forced on you by the single namespace. it is just convenient for the packager.
|
|
#
?
Nov 30, 2014 23:09
|
|
|
Yeah, I'm not the biggest fan of Poettering's proposal for a variety of other reasons as well. I think the locked down SDK will make for a lovely development experience. I started working on my own project before realizing that would just add more fragmentation to the problem, so I wrote a bunch of notes which I'll talk to Poettering about in person.
|
|
#
?
Nov 30, 2014 23:09
|
|
|
poettering's proposal is a big bag of hurt and stupid. i can't touch that right now. your essay is a different animal
|
|
#
?
Nov 30, 2014 23:10
|
|
|
Suspicious Dish posted:The issue is with "they build one filesystem tree" is that you can't upgrade apps and your OS independently. You have to update in lockstep, or you fail to upgrade both. This is because ELF symbol resolution is dumb and shared library authors are incompetent.
|
|
#
?
Nov 30, 2014 23:11
|
|
|
pseudorandom name posted:This is because ELF symbol resolution is dumb and shared library authors are incompetent. solaris did ELF right: rpath is mandatory linux's global linker configuration is easier for idiots to deal with, but it's bad and broken behavior just to put some wildcards in a file and say "yeah just get my symbols out of that cesspit of libraries, it'll be fine"
|
|
#
?
Nov 30, 2014 23:14
|
|
|
Here's how to atomically upgrade /usr: 1. Create a /.usr.sysupgrade folder 2. Hardlink every existing file in /usr into a replicated /.usr.sysupgrade directory structure 3. Upgrade poo poo in /.usr.sysupgrade . Do the same thing for /etc as well if needs be: this will be easier when dropfiles become a more common pattern, because that frees you from having to do the package manager diff-merge poo poo you described 4. Use a new syscall flip() or whatever to atomically swap /.usr.sysupgrade and /usr 6. Delete /.usr.sysupgrade If you use ext4 like a sensible person then your FS metadata is serialized and journalled, so you can always recover from this after a crash by deleting /.usr.sysupgrade (e: never mind the poo poo I just posted about some extra flag file, this is unnecessary, but this is just the simplest poo poo I came up with in like 2 minutes) No btrfs bullshit required. Sapozhnik fucked around with this message at 23:18 on Nov 30, 2014 |
|
#
?
Nov 30, 2014 23:15
|
|
|
Notorious b.s.d. posted:solaris oh you
|
|
#
?
Nov 30, 2014 23:15
|
|
|
for the record i am not so sure that upgrading the OS between releases is a reasonable design goal for a package system i get that desktop users love it, but that doesn't make it right
|
|
#
?
Nov 30, 2014 23:18
|
|
|
I think it's a goal that any sane system management tool should strive to support. You're absolutely right that package managers can't handle it.
|
|
#
?
Nov 30, 2014 23:22
|
|
|
the sane way to manage a system upgrade is to deploy a new one with a separate IP endpoint and LB pool, run a health check, and delete the old one if it passes pls make this work on my desktop tia
|
|
#
?
Nov 30, 2014 23:28
|
|
|
On other platforms, upgrading the OS carries no guarantees that third party software will continue to work
|
|
#
?
Nov 30, 2014 23:31
|
|
|
Notorious b.s.d. posted:solaris did ELF right: rpath is mandatory rpath wouldn't help for the real-world linux problems of "two incompatible versions of the library are loaded into the same linkmap" and "gently caress it we broke everything on this minor point release you get what you pay for"
|
|
#
?
Dec 1, 2014 00:21
|
|
|
linux shared library authors are too stupid to even do the most basic of things like pass --default-symver to ld
|
|
#
?
Dec 1, 2014 00:23
|
|
|
Captain Foo posted:Aren't apps workable because the runtime is guaranteed, where this isn't true across Linux distros? the fundamental thing that OS X and iOS guarantee is release-to-release binary compatibility, with managed deprecation, when upgrading the base OS. there is also extensive work done to ensure both backward and forward compatibility of software running on these platforms: - existing software built for an earlier OS version may still get older behaviors from the new OS because it can tell what framework & library versions it was build against. - new software can be built to be compatible across a range of OS versions, and implement graceful fallback behavior when running on an older OS that doesn't have some newer API. does any Linux even attempt to manage things like API set availability over major or minor OS versions?
|
|
#
?
Dec 1, 2014 02:49
|
|
|
No, the only part of Linux that even approaches that is glibc, and its solution to supporting old versions is "build on that old version" the majority of the Linux software stack is built by idiots using broken tools and all the things that exist to allow you to support ABI versioning and evolution go completely unused pseudorandom name fucked around with this message at 02:56 on Dec 1, 2014 |
|
#
?
Dec 1, 2014 02:54
|
|
|
Captain Foo posted:On other platforms, upgrading the OS carries no guarantees that third party software will continue to work it sure does on OS X and iOS, as long as that software consists of apps built in the standard fashion against the OS-supplied APIs
|
|
#
?
Dec 1, 2014 03:18
|
|
|
pseudorandom name posted:the majority of the Linux software stack is built by idiots using broken tools and all the things that exist to allow you to support ABI versioning and evolution go completely unused it took a few years to get OS X developers all building against SDKs containing headers and stub libraries (via -isysroot) rather than /, and to get them used to setting the Deployment Target (minimum OS on which to run) separately from the SDK (maximum OS from which to use features). on Linux even trying to nail down what "OS version" means will cause years of debate and flaming and forking and petitions.
|
|
#
?
Dec 1, 2014 03:23
|
|
|
I tried out gnome 3.14 on fedora 21 today, it was real nice and I could even tether my iPhone and connect to our Cisco vpn
|
|
#
?
Dec 1, 2014 10:42
|
|
|
cons: gnome shell currently runs at 60-80% cup usage at all times
|
|
#
?
Dec 1, 2014 10:42
|
|
|
What graphics chip / driver are you using?
|
|
#
?
Dec 1, 2014 16:49
|
|
|
off to a new jorb in which i fully expect to use a linux on a desktop assuming i have the ability to install my own (i don't know yet, old jorb i ended up being my own unix sysadmin) is there a good reason to use fedora over rhel/centos 7? engineering tools always support rhel exclusively but if fedora isn't wildly incompatible and offers benefits over rhel i might try it
|
|
#
?
Dec 1, 2014 18:56
|
|
|
BobHoward posted:off to a new jorb in which i fully expect to use a linux on a desktop you'll probably have a much easier time getting newer rpms for fedora
|
|
#
?
Dec 1, 2014 18:58
|
|
|
they call it centos because it takes a century to get updates 
|
|
#
?
Dec 1, 2014 19:01
|
|
|
use oracle linux
|
|
#
?
Dec 1, 2014 20:23
|
|
|
Suspicious Dish posted:What graphics chip / driver are you using? virtual box and vboxdriver with guest additions and all that enabled 3d support enabled because gnome shell won't start with it disabled
|
|
#
?
Dec 1, 2014 22:44
|
|
|
fyi 95% of all systemd bitching could be eliminated if the loving fucks just did the whole awesome journaled thing but still wrote logs out to their normal locations while we learn the differences i had to google where the gently caress xorg.0.log went and thats some poo poo i havent had to do since gentoo
|
|
#
?
Dec 1, 2014 22:53
|
|
|
|
| # ? May 23, 2024 19:44 |
|
|
ahmeni posted:fyi 95% of all systemd bitching could be eliminated if the loving fucks just did the whole awesome journaled thing but still wrote logs out to their normal locations while we learn the differences Does this sound like the normal way Linux does things? And who is it being pushed hard by? Red Hat. Okay, so where does Red Hat get THEIR finances from? Three-letter agencies: DoD, FBI, NSA, CIA. In fact, more than 85 percent of Red Hat's money comes either directly or indirectly from the U.S. government. So, my question is this: In light of Snowden and Manning, why is nobody raising a red flag at this out-of-the-blue atypical behavior coming from the top tier of the Linux distros when it's being pushed by a corporation practically in the pocket of the USA security machine? If you don't smell something hinky when Linux distros out of the blue adopt an "our way or the highway" while burying threads and banning users - and it all comes from a company that literally cashes checks from the NSA - I'm sorry, but this smells fishier than a tuna factory in August.
|
|
#
?
Dec 2, 2014 02:34
|
|