You realise one of the key reasons for adopting systemd was that distribution maintainers have LESS work to do right?
That's the only upside of systemd I know of: it reduces the work of maintainers if they ship upstream integration as-is. But if the maintainers try to improve it, it all falls apart (case in point: Debian systemd maintainers still didn't manage to split the package to put the kitchen sink, bicycle and fish bowl (aka different components of systemd) apart. As for benefits for the user... nope. But alas, when distribution maintainers and users disagree, the former prevail.
Not having to manage a shitload of nasty scripts was one of its great selling points to the maintainers.
Right. A typical init script is one line (using #!/lib/init/init-d-script), systemd usually requires you to edit 3-5 files.
Nope - my best idea for any IoT devices that require connection to a vendor server is to hit them repeatedly with a hammer.
What about this: any IoT device should refuse to contact the wide Internet unless it can periodically contact an user-configurable update server?
This would handle all major use cases: 1. no network, 2. local network only, 3. Internet at large; provide a reasonable default for the uneducated crowd while giving control to those who want it, and provide a configurable compromise between privacy and updates.
Ubuntu uses apt, and there's a large selection of tools to set up your own mirrors, caches or own repositories. I for one prefer apt-cacher-ng and reprepro for my home usage, but there's more than ten tools in either category I can name out of the top of my head.
OK, so perhaps not "constructive" in the literal sense, but still...
Technically "deconstructive", but whacking misbehaving vendors with a hammer just can't go wrong.
This doesn't seem to be an attack on ProtonMail directly, but an attempt to reduce their userbase. Even a skilled rational user will spend a bit of time researching these claims, while the paranoid but dumb crowd are going to jump ship to some snake oil "secure" providers.
I found a claim that new monitors started doing this around 2005, and that (post from 2013) "the majority of monitor sales" do this. So, it's good to have on real computers, too.
I'd thus like to advertise a nice GTK2+3 theme: you want to "apt install darkcold-gtk-theme" (or darkmint if you prefer green). Despite no css skills I fixed them up, and since then the original author returned from years of no show. Somehow, any other theme shipped in Debian is at most mid-gray rather than actual black.
It's not about XFS itself failing, it's about the disk, its firmware, cable (@#$%^&*!) or some part of the mobo silently passing corrupt data as valid. XFS has no way to detect this.
For this reason ext4 recently got metadata checksums as well, and it's a MASSIVE, vital improvement. Sure, it has only a small chance of catching silent corruption, but as such errors rarely go alone, you at least know you have a problem. Unlike btrfs, you won't know of _data_ corruption nor know which blocks are bad, but that's still a great thing to have.
As for LVM: it would need to have a large amount of unallocated space somewhere. The filesystem also needs a large mount of unallocated space. You also can't resize XFS down (at all) nor ext4 (offline only) so you can't even move from one pool to the other (discard somewhat helps but you still risk overcommit). Btrfs and ZFS on the other hand, violating that layer, combine that pool into one and can utilize it fully.
And, LVM can't work with new technologies such as DIMM-mounted Optane. Ok, it can work by emulating a legacy disk, but you lose all the goodies. I see no obvious way to port btrfs or Stratis to DAX so both will be obsolete. The replacement functionality is done by the CPU itself, and CoW is done mostly as a page fault rather than a full filesystem operation -- you can't really hide it in a layer. Heck, going through the kernel at all causes such a relative slowdown that you want to go all userspace:)
Stratis, no matter its true or perceived upsides, is simply too late. Btrfs is fully functional today, and Optane memory is being shipped to first rollouts. The marketing bleating currently goes "your datacenter storage will get XXX times faster" rather than "forget game loading times or pauses to save", but the technology is here, live.
To elaborate: on-disk data gets corrupted both at rest and during transit. They _are_ supposed to report errors, but they notoriously fail to do so -- then the pipeline is notoriously flaky. Companies I worked with tend to buy bottom of the barrel hardware, and in my experience silent data corruption happens drastically more often with disks than memory. Thus, btrfs is needed for the same reason you use ECC RAM for, except that its error-spotting gets a lot more use than that of RAM.
Things may change in the future when $DISK gets far closer (such as DIMM-interface Optane), but that doesn't work with CoWing LVM anyway (well it does, but you lose so much speed you could have stayed with traditional disks). So that'll obsolete both btrfs and Stratis; for now I use what's good for current hardware. And btrfs is here, Stratis is not.
That report is positively ancient, and no one should look at particular approaches believed most promising 42 years ago. But it made a particular promise about research speed vs funding, and this promise turned out true. So the report is still valid to refute complaints that "fusion is always 30 years in the future".
Funny that -- I'd never trust any silentdatalossfs like XFS on a server -- or desktop, or a SoC -- either. There's a massive difference between knowing your data has been corrupted months after the fact and during the next read (or scrub). And, with my rate of mistakes, hourly backups without having to stat() every file is also a massive ass-saver.
Snapshot capabilities of LVM are sharply limited, and getting them feature parity with btrfs would be effectively reinventing btrfs. There's no way around CoW, merely different possible implementations -- or different layers doing it.
I believe $DISK's FTL would be the best place, as it _already_ does CoW, but unless it starts exposing such capabilities to the kernel...
I guess I am confused on why btrfs is being deprecated. Any ideas why?
Business reason: Red Hat is making its own storage system that poorly reinvents parts of btrfs in a higher layer, touted as enterprisey with paid features.
Technical reason: Red Hat uses ridiculously ancient kernels, backporting features from kernels 30 or so versions newer. This just can't be stable unless you have a team of engineers devoted to every subsystem, and Red Hat never had such a team for btrfs. The kernel moves quickly, without regards to internal compat, such backporting just can't go well.
Don't you replace pants from time to time, and toss away or donate (damaged/fine) old ones when they are no longer good for you?
Weight is not volatile enough for this to be a problem. If you don't have your own data, you can look at mine. I did rapidly lose 9kg recently, and all it meant was belt no longer being optional with old pants, a single hole tighter.
I gotta say, the ones that leave voicemails, filling up my mailbox, are the worst!
Can't you just let it fill up and leave it at that? I have yet to see a human listen to, or intentionally record, a voicemail.
For some reason phone companies don't let you disable voicemail or make it hard to do so, but since you (at least here) don't pay for storage, you can just ignore it.
I still don't get why anybody puts up with Gnome when KDE is available.
Heck, even when twm is available. Anything available in that field is better than Gnome3. It's the epitome of what "UX designers" stand for. There are two valid directions for an UI to go for: ease for new users, and ergonomy for advanced ones. Gnome3 blows both to a ridiculous degree, at the same time trashing efficiency, portability (works on x86 only), and so on.
I really don't get why Ubuntu ships with Gnome standard.
The ranking in TFA mixes companies and countries. If you look at just the latter, you see: 1 China (Coal) 14.32% 6 Coal India 1.87% 8 Russia (Coal) 1.86% 15 Poland Coal 1.16% So we're 4th, beaten only by [sub-]continent spanning major countries, despite ours population of mere 38.5M.
All greenhouse gas reduction activity was not only stopped but even reversed by our glorious National Communist government: they actually open new mines and coal power plants, and made some forms of better energy generation basically illegal (like, "quiet zones" required around any new or modernised wind generators mean you can't put them pretty much anywhere).
Getting a high place in a per-population contest isn't hard, doing "well" in absolute numbers when compared to much more populous countries is quite an accomplishment. So our "Good Change" regime did make Poland a "leading country" after all!
So you say it's better to allow advertising and taking billions of revenue for make-believe treatments without any validation whatsoever? This is why we have homeopathy and the like.
An "ethics committee" would fail its job if it allowed such a treatment to go forward (beyond a clearly marked experimental stage), not if it demanded validation.
Traditional humanities majors can. Gender studies majors can't
There's a big difference between a rational proto-science (heck, logic originated from philosophy!), quasi-science or even a sane description of something inherently non-scientific (like literature) -- and Orvellian doublethink that teaches people something contrary to obvious observation.
The former group teaches students a way of thinking. The latter group is religion.
Lack of replication is often cited as evidence that science is bunk
Because, for replicatable studies, such so-called-science is indeed bunk. You can replicate psychological studies, and for any kind of treatments, you can have control groups, preferably blind. So that "behavioural therapy" of yours can be tested. And that it can but its proponents failed to do so is exactly why we call bullshit.
wat? That'd be a ridiculous layering violation. The core rule of writing reliable tools is: "do one thing and do it well". A communication medium needs a system for managing files about as much as a fish needs a bicycle.
What else would you propose, that an init system should include a web server and parse untrusted XML in it?
Slashdot Mirror
You realise one of the key reasons for adopting systemd was that distribution maintainers have LESS work to do right?
That's the only upside of systemd I know of: it reduces the work of maintainers if they ship upstream integration as-is. But if the maintainers try to improve it, it all falls apart (case in point: Debian systemd maintainers still didn't manage to split the package to put the kitchen sink, bicycle and fish bowl (aka different components of systemd) apart. As for benefits for the user... nope. But alas, when distribution maintainers and users disagree, the former prevail.
Not having to manage a shitload of nasty scripts was one of its great selling points to the maintainers.
Right. A typical init script is one line (using #!/lib/init/init-d-script), systemd usually requires you to edit 3-5 files.
But will linux on Windows 10 support systemd?
No.
Nope - my best idea for any IoT devices that require connection to a vendor server is to hit them repeatedly with a hammer.
What about this: any IoT device should refuse to contact the wide Internet unless it can periodically contact an user-configurable update server?
This would handle all major use cases: 1. no network, 2. local network only, 3. Internet at large; provide a reasonable default for the uneducated crowd while giving control to those who want it, and provide a configurable compromise between privacy and updates.
Ubuntu uses apt, and there's a large selection of tools to set up your own mirrors, caches or own repositories. I for one prefer apt-cacher-ng and reprepro for my home usage, but there's more than ten tools in either category I can name out of the top of my head.
OK, so perhaps not "constructive" in the literal sense, but still...
Technically "deconstructive", but whacking misbehaving vendors with a hammer just can't go wrong.
This doesn't seem to be an attack on ProtonMail directly, but an attempt to reduce their userbase. Even a skilled rational user will spend a bit of time researching these claims, while the paranoid but dumb crowd are going to jump ship to some snake oil "secure" providers.
Unless you're using an OLED display, or one with local dimmin, a dark theme does nothing to save energy
Right then. A dark theme is still good for your eyes during the night, so you save wetware even if no electricity.
I found a claim that new monitors started doing this around 2005, and that (post from 2013) "the majority of monitor sales" do this. So, it's good to have on real computers, too.
I'd thus like to advertise a nice GTK2+3 theme: you want to "apt install darkcold-gtk-theme" (or darkmint if you prefer green). Despite no css skills I fixed them up, and since then the original author returned from years of no show. Somehow, any other theme shipped in Debian is at most mid-gray rather than actual black.
It's not about XFS itself failing, it's about the disk, its firmware, cable (@#$%^&*!) or some part of the mobo silently passing corrupt data as valid. XFS has no way to detect this.
For this reason ext4 recently got metadata checksums as well, and it's a MASSIVE, vital improvement. Sure, it has only a small chance of catching silent corruption, but as such errors rarely go alone, you at least know you have a problem. Unlike btrfs, you won't know of _data_ corruption nor know which blocks are bad, but that's still a great thing to have.
As for LVM: it would need to have a large amount of unallocated space somewhere. The filesystem also needs a large mount of unallocated space. You also can't resize XFS down (at all) nor ext4 (offline only) so you can't even move from one pool to the other (discard somewhat helps but you still risk overcommit). Btrfs and ZFS on the other hand, violating that layer, combine that pool into one and can utilize it fully.
And, LVM can't work with new technologies such as DIMM-mounted Optane. Ok, it can work by emulating a legacy disk, but you lose all the goodies. I see no obvious way to port btrfs or Stratis to DAX so both will be obsolete. The replacement functionality is done by the CPU itself, and CoW is done mostly as a page fault rather than a full filesystem operation -- you can't really hide it in a layer. Heck, going through the kernel at all causes such a relative slowdown that you want to go all userspace :)
Stratis, no matter its true or perceived upsides, is simply too late. Btrfs is fully functional today, and Optane memory is being shipped to first rollouts. The marketing bleating currently goes "your datacenter storage will get XXX times faster" rather than "forget game loading times or pauses to save", but the technology is here, live.
in the ads that I see
Found your problem right here.
To elaborate: on-disk data gets corrupted both at rest and during transit. They _are_ supposed to report errors, but they notoriously fail to do so -- then the pipeline is notoriously flaky. Companies I worked with tend to buy bottom of the barrel hardware, and in my experience silent data corruption happens drastically more often with disks than memory. Thus, btrfs is needed for the same reason you use ECC RAM for, except that its error-spotting gets a lot more use than that of RAM.
Things may change in the future when $DISK gets far closer (such as DIMM-interface Optane), but that doesn't work with CoWing LVM anyway (well it does, but you lose so much speed you could have stayed with traditional disks). So that'll obsolete both btrfs and Stratis; for now I use what's good for current hardware. And btrfs is here, Stratis is not.
Sounds like they tested their product -- a lot, and that their business plan involves giving prospective buyers a test drive.
That report is positively ancient, and no one should look at particular approaches believed most promising 42 years ago. But it made a particular promise about research speed vs funding, and this promise turned out true. So the report is still valid to refute complaints that "fusion is always 30 years in the future".
There's no way I'd trust it on a server
Funny that -- I'd never trust any silentdatalossfs like XFS on a server -- or desktop, or a SoC -- either. There's a massive difference between knowing your data has been corrupted months after the fact and during the next read (or scrub). And, with my rate of mistakes, hourly backups without having to stat() every file is also a massive ass-saver.
Snapshot capabilities of LVM are sharply limited, and getting them feature parity with btrfs would be effectively reinventing btrfs. There's no way around CoW, merely different possible implementations -- or different layers doing it.
I believe $DISK's FTL would be the best place, as it _already_ does CoW, but unless it starts exposing such capabilities to the kernel...
The explanation is simple.
I guess I am confused on why btrfs is being deprecated. Any ideas why?
Business reason: Red Hat is making its own storage system that poorly reinvents parts of btrfs in a higher layer, touted as enterprisey with paid features.
Technical reason: Red Hat uses ridiculously ancient kernels, backporting features from kernels 30 or so versions newer. This just can't be stable unless you have a team of engineers devoted to every subsystem, and Red Hat never had such a team for btrfs. The kernel moves quickly, without regards to internal compat, such backporting just can't go well.
So, sadly, I"M guessing IBM will acquire and fuck up RHEL
Uhm, Gnome3, systemd, NetworkManager? Or, have you seen people try to upgrade Red Hat boxes? The only way for RHEL is up.
and my pants fit again
Don't you replace pants from time to time, and toss away or donate (damaged/fine) old ones when they are no longer good for you?
Weight is not volatile enough for this to be a problem. If you don't have your own data, you can look at mine. I did rapidly lose 9kg recently, and all it meant was belt no longer being optional with old pants, a single hole tighter.
LMDDGTFY. Already done, log ago.
I gotta say, the ones that leave voicemails, filling up my mailbox, are the worst!
Can't you just let it fill up and leave it at that? I have yet to see a human listen to, or intentionally record, a voicemail.
For some reason phone companies don't let you disable voicemail or make it hard to do so, but since you (at least here) don't pay for storage, you can just ignore it.
I still don't get why anybody puts up with Gnome when KDE is available.
Heck, even when twm is available. Anything available in that field is better than Gnome3. It's the epitome of what "UX designers" stand for. There are two valid directions for an UI to go for: ease for new users, and ergonomy for advanced ones. Gnome3 blows both to a ridiculous degree, at the same time trashing efficiency, portability (works on x86 only), and so on.
I really don't get why Ubuntu ships with Gnome standard.
Nor do I. It's the systemd of window managers.
The ranking in TFA mixes companies and countries. If you look at just the latter, you see:
1 China (Coal) 14.32%
6 Coal India 1.87%
8 Russia (Coal) 1.86%
15 Poland Coal 1.16%
So we're 4th, beaten only by [sub-]continent spanning major countries, despite ours population of mere 38.5M.
All greenhouse gas reduction activity was not only stopped but even reversed by our glorious National Communist government: they actually open new mines and coal power plants, and made some forms of better energy generation basically illegal (like, "quiet zones" required around any new or modernised wind generators mean you can't put them pretty much anywhere).
Getting a high place in a per-population contest isn't hard, doing "well" in absolute numbers when compared to much more populous countries is quite an accomplishment. So our "Good Change" regime did make Poland a "leading country" after all!
So you say it's better to allow advertising and taking billions of revenue for make-believe treatments without any validation whatsoever? This is why we have homeopathy and the like.
An "ethics committee" would fail its job if it allowed such a treatment to go forward (beyond a clearly marked experimental stage), not if it demanded validation.
Traditional humanities majors can. Gender studies majors can't
There's a big difference between a rational proto-science (heck, logic originated from philosophy!), quasi-science or even a sane description of something inherently non-scientific (like literature) -- and Orvellian doublethink that teaches people something contrary to obvious observation.
The former group teaches students a way of thinking. The latter group is religion.
Lack of replication is often cited as evidence that science is bunk
Because, for replicatable studies, such so-called-science is indeed bunk. You can replicate psychological studies, and for any kind of treatments, you can have control groups, preferably blind. So that "behavioural therapy" of yours can be tested. And that it can but its proponents failed to do so is exactly why we call bullshit.
wat? That'd be a ridiculous layering violation. The core rule of writing reliable tools is: "do one thing and do it well". A communication medium needs a system for managing files about as much as a fish needs a bicycle.
What else would you propose, that an init system should include a web server and parse untrusted XML in it?
No spying.
You mean, like Google says about Gmail, despite evidence to the contrary?