The US Marine Corp used horses during the Korean War. Famously:
"She served in numerous combat actions during the Korean War, carrying supplies and ammunition, and was also used to evacuate wounded. Learning each supply route after only a couple of trips, she often traveled to deliver supplies to the troops on her own, without benefit of a handler. The highlight of her nine-month military career came in late March 1953 during the Battle for Outpost Vegas when, in a single day, she made 51 solo trips to resupply multiple front line units. She was wounded in combat twice." https://en.wikipedia.org/wiki/...
The Marines still train to use pack animals in mountain warfare. I believe the US Army also has some training in the use of pack animals. I'm not sure if it is just for Special Forces or if "regular" Mountain Warfare units also have the training/capability.
Just to illustrate: "You say that like an Air Force Puke, I suppose you also think the F35 can replace the A10 too!" My guess, is he doesn't really know the difference between A and F aircraft, or else he wouldn't have conflated the roles like he did.
Actually the F-35 is considered a multimission aircraft, both a fighter and an attack aircraft. Hence it being referred to as the Joint Strike Fighter, "Strike" as in ground attack. And the Pentagon in fact says the F-35 is the "replacement" for the A-10, that the F-35 will provide close air support. And of course this is a controversial plan.
The A-10 is an anomaly, as is the original F-16, both being single purpose aircraft, and both originating from rogue teams working outside the normal Pentagon procurement process. The original F-16 could drop bombs but that capability was integrated with minimal compromises to air-to-air capability. Unlike a more traditional multimission aircraft where the tradeoffs are more dramatic. Later model F-16s moved a bit more toward multimission. The F-15 and F-14 were multimission from day one, as was the F-4 before them. Only the F/A-18 is truthful in its naming, "advertising" its multimission nature. Basically for decades if it originated from the Pentagon it was multimission.
Oh, and of course there is the F-117 which is in no way a fighter, pure ground attack. But was labeled as a fighter because they wanted to recruit fighter pilots for the program and to mislead foreign intelligence services.
In other words the use of "F" and "A" is very complicated.
Don't get stuck on today's missile designs. I expect the future AI enabled missiles to be larger, have larger control surfaces and a variable output engine so that they can go fast or maneuver as needed. In short they will be somewhat like a miniature aircraft. Think something "closer" to a cruise missile in look than a sparrow or sidewinder. Hmm... those are more of yesterday's designs rather than todays but hopefully the point comes across.
My Dad too, that's why I'm familiar with the local Guard unit's history.
Every GOOD scout knows his greatest and most effective weapon... the RADIO.
Same as in my Dad's day many decades ago. That's why I wrote "A role not unlike that 1930s horse cavalry role, eyes, not direct combat." The radio is an indirect sort of combat. Basically, in my Dad's day and probably yours if the scout is shooting the scout is probably "seen", and you've described the problem with being seen.
The big advantage of a pilot over a drone is that you can't jam or spoof a pilot. Same reason that wire-guided ATGMs remain a major player in modern battlefields.
The missiles I am suggesting will be human targeted and launched but once launched they will be self guided. Their sensors should be more capable than a human eyeball v1.0. Which is sort of what they currently do, but with better sensors and a more variable engine allowing for greater maneuverability, and an on-board AI to employ these, it will be a game changer. A fast stealthy aircraft is still useful as a launch platform, like a B-52 is today with cruise missiles, but the days of high maneuverability manned flight are drawing to a close. Again, in an air-to-air sense, for recon the current "fighter" capabilities will probably persist a while longer.
Electronic gunsights adjusting for target range and speed, telling the pilot when/where to shoot, date back to the 1950s and the Korean war. Not as fancy as a modern, or even somewhat modern hud, but the essentials are old tech.
If you can't see the difference between a poorly-maneuvering missile announcing its location from miles away and a fast jet actively engaging the target from within AAM range, your last sentence might apply to you more than me.
If you think missiles maneuver poorly compared to fighters, or are more detectable, you are indulging in the 420 too much. The only advantage fighters have over missiles are range. However even 1970s cruise missile technology has medium range low level terrain masking flight capabilities. They can be launched by bombers, tomorrow smaller versions will be launched by fighters. The AI will go into the missiles not the fighters. The fighters will become more of a launch platform for such missiles, like the B-52 launching cruise missiles today. Range and stealth will remain important, but high G maneuvering will not.
Missiles only seem low maneuverable, sometimes fighters avoiding one, because of their high speed. This high speed is needed to overtake a fleeing fighter. The high speed translates to a higher turning radius. Give missiles a propulsion system that is more variable and fighters will have little chance to outmaneuver them. Don't let the limitations of current technology confuse you.
As you say the Norden allowed more accurate bombing from high altitude but it was never anything near precise, even by WW2 standards. In WW2 when you wanted precision, in the sense that you hit a particular building, you sent in fighters or medium bombers (B-25, B-26) at low altitude. Heavy bombers (B-17, B-24) at altitude only hit a particular building through large numbers of bombs and statistical averages even with the Norden. The heavies were also more precise at low altitude but more vulnerable than the mediums. By the end of the war the heavies were found to be most effective at incendiaries where precision was not required.
That said, the Norden was an absolutely amazing piece of technology for its day.
The F35 and the A10, while both are "fighter" jets, have completely different roles. The F35 is for establishing air dominance so that the A10 can ground n pound.
Hopefully you are using "fighter" in quotes to really mean combat aircraft. But no, the A-10 is not a fighter. Its a ground attack aircraft. It carries air-to-air missiles as a last resort defense. Sort of like a member of a tank crew carrying a pistol. And like the tank crew member employing a pistol, if the A-10 pilot is employing air-to-air missiles something has gone terribly terribly wrong.
A lot of people overinterpret the lessons of the Korean war where missiles were overstressed versus the technology of the time... and have taken it as some universal lesson which will apply forever into the future, that close-range dogfighting will always be the most critical aspect of aircraft design.
Vietnam not Korea. Personally I expect the AI to go into the missiles not the aircraft. Fighters becoming a romantic anachronism, like horse cavalry. And like horse cavalry they will last longer than people expect. My local National Guard unit is cavalry, reconnaissance, and had horse as late as the 1930s. In certain terrain guys sneaking around on horse was still more effective than vehicles. They were just the eyes for armored formations and not expect to fight themselves. Sort of like modern Long Range Reconnaissance Patrols, if you are firing your weapons something has gone terribly wrong. Note some US Special Forces briefly operated as cavalry in Afghanistan. I believe the US Marines sometimes use dirt bikes. Recon may also be a role for repurposed fighters. Actually it has been such a role, removing guns an armor and adding cameras. Sometime there are gaps with satellite and drone coverage and a fast mover flying low and masking its approach with terrain fills that gap. A role not unlike that 1930s horse cavalry role, eyes, not direct combat.
I've yet to see any form of public transit without windows. People wouldn't ride it.
Windows will become virtual. Your display can show you a movie, news, stock charts, etc or the outside world. And to be honest, imho, if the windows don't open to let in some air then virtual might not be that bad. Well, assuming, you are only a passenger and are not expected to take control of the vehicle at some point.
I know nothing about Chinese patent law, other than that WTO membership requires China to have one. But in the United States, the switch to "first to file" did not abridge the novelty requirement. Anything published, even other than as a patent application, before a given patent application is filed is considered prior art and therefore "filed".
No one said prior art no longer exists. The point is that Baili shipping first is not necessarily meaningful. If Apple filed before Baili shipped, and or course before Baili filed, then Apple "wins". Apple shipping second is not necessarily a problem as long as Apple filed early enough.
I think the patent system in China and in the US (now) is First to File. Discovering first or shipping first does not matter. So all that matters is the dates of the patent applications.
View All shows 2011-2016 for E7, 2012-2016 for E5, but only 2011-2012 for E3. Apparently the E3 View All page is broken, apologies. The normal E3 landing page has an incomplete list of processors. It seems only the Compare Product Families page gives a complete list.
I did a BYO PC recently with an i7 gen 6. The comparable E3 v5 seem to either run hotter when the price is comparable or better, or are much more expensive when the they run with the same thermal output. Both situations would seem troublesome for Apple.
For you it may be a low risk. For Apple its not. Apple will be shipping millions of machines.
And these machines are already vulnerable just to single bit errors anywhere both in the IO path and in memory. The repair-of-death you describe involves multiple errors...
And that is how airplanes occasionally crash. Its usually not one flaw or problem, its multiple problems/flaws occurring at the same time. If you are an individual wishing to fly this is of little concern, the odds are with you. If you are an aircraft manufacturer it is of concern since your aircraft will conduct millions of flights and there are hostile lawyers eyeing your deep pockets.
Apple was in a similar situation, but with deeper pockets to attract class action lawyers.
Stop approaching this from the perspective that ZFS is flawed. Rather approach this from the perspective that ZFS assumes memory can be trusted
... so does every other filesystem. I'll quote another bit of that paper you like:
There is quite a difference between corrupting the inode info / timestamp info and corrupting the **contents of a file**, its user data. That is what is unique about ZFS. File data being **read** is at risk due to automatic repairs of **user data**.
Intel is more likely to support ECC in lower end CPUs (ex i3) than in mid to higher end CPUS (ex i5, i7)
i7-class Xeons (E3-XXXX) support ECC and are usually priced basically identically to their i7 cousins.
Yes, server grade CPUs support server grade RAM. And judging from Intel's data sheets the current generation Xeons are slower (clock rate, more cores though) and generate more heat. Notice Apple's consumer products are designed for more moderate thermal conditions. And the 4-5 year old Xeons you mention seem to be slightly more expensive that the new i7s in Intel's price lists.
Another difficulty for a consumer oriented company like Apple, making using ECC not really an option for them
I'm sure Apple are more than capable of pushing for it if they considered it a priority. They have the purchasing power, they have the margins, they have the PR to make people wet themselves over the benefits if they so choose.
Intel used to support ECC in some i5 and i7, if I'm reading the 6th gen data sheets correctly no i5 or i7 supports ECC. I severely doubt Apple can change Intel's direction. Plus, no, consumers won't care so why would Apple even bother? ZFS was interesting when they offered servers, they no longer do so, the interest is gone, its not their mission to advocate for ZFS.
The bit flips required to make such a process actually damage your data seems quite convoluted - it'd have to be multiple errors in different locations happening at just the right times...
For you it may be a low risk. For Apple its not. Apple will be shipping millions of machines. Plus given prior knowledge of ZFS' unique failure mode they will be vulnerable in court, a hostile class action lawyer arguing they increased their customer's vulnerability through their choice of filesystem and type of RAM.
Is ZFS really significantly more vulnerable to this by comparison, or is it just that ZFS won't defend you against it?
More vulnerable. The important distinction is that ZFS can corrupt on-disk data when applications are performing read operations. That is what differentiates ZFS from other filesystems.
Stop approaching this from the perspective that ZFS is flawed. Rather approach this from the perspective that ZFS assumes memory can be trusted and should be used with system equipped with ECC RAM. Every piece of software has system requirements. ZFS just has an extra one. Given ZFS' capabilities it an excellent tradeoff, but its not a tradeoff available to consumer hardware.
I'm going the BYO route for my SOHO server, as I have been doing for PCs for decades. Motherboard, CPU and RAM all supporting ECC. For those reading along you need all three, and strangely Intel is more likely to support ECC in lower end CPUs (ex i3) than in mid to higher end CPUS (ex i5, i7). Another difficulty for a consumer oriented company like Apple, making using ECC not really an option for them. ZFS is just not a good fit for Apple now that they no longer sell server grade equipment, hence their losing interest.
No, ZFS is more vulnerable. All filesystems can corrupt the disk by writing bad data, however ZFS can corrupt data on disk when an application only reads the disk. When checksums fail ZFS will assume the problem is on disk and attempt to "repair" the data on disk. This automatic repair is a great feature, when your RAM can be trusted.
One example from the cited paper: "Dirtying blocks due to updating file access time increases the possibility of making corruptions permanent. By default, access time updates are enabled in ZFS; therefore, a read-only workload will update the access time of any file accessed. Consequently, when the structure containing the access time (znode) goes inactive (or when there is another workload that updates the znode), ZFS writes the block holding the znode to disk and updates and writes all its parental blocks. Therefore, any corruption to these blocks will become permanent after the flush caused by the access time update."
I am not saying it is a new risk in ZFS, merely that it is something understood about ZFS for a long time. And given that Apple is making consumer devices and does not want to use ECC, ZFS is not a good fit and that is most likely why Apple passed on it. If Apple were still making servers then ZFS may very well be an option on those.
You said "ZFS needs ECC RAM". Implying that ZFS is special is this regard. The link however contains the following:
7 Beyond ZFS
In addition to ZFS, we have applied the same fault injec- tion framework used in Section 5 to a simpler filesystem, ext2. Our initial results indicate that ext2 is also vulner- able to memory corruptions. For example, corrupt data can be returned to the user or written to disk. When cer- tain fields of a VFS inode are corrupted, operations on that inode fail or the whole system crashes. If the inode is dirty, the corrupted fields of the VFS inode are propa- gated to the inode in the page cache and are then written to disk, making the corruptions permanent. Moreover, if the superblock in the page cache is corrupted and flushed to disk, it might result in an unmountable filesystem.
In summary, so far we have studied two extremes: ZFS, a complex filesystem with many techniques to maintain on-disk data integrity, and ext2, a simpler filesystem with few mechanisms to provide extra relia- bility. Both are vulnerable to memory corruptions. It seems that regardless of the complexity of the file sys- tem and the amount of machinery used to protect against disk corruptions, memory corruptions are still a problem.
ZFS needs ECC as much as every other file system.
No, ZFS is more vulnerable. All filesystems can corrupt the disk by writing bad data, however ZFS can corrupt data on disk when an application only reads the disk. When checksums fail ZFS will assume the problem is on disk and attempt to "repair" the data on disk. This automatic repair is a great feature, when your RAM can be trusted.
Ex: "Dirtying blocks due to updating file access time increases the possibility of making corruptions permanent. By default, access time updates are en- abled in ZFS; therefore, a read-only workload will update the access time of any file accessed. Consequently, when the structure containing the access time (znode) goes inactive (or when there is another workload that updates the znode), ZFS writes the block holding the znode to disk and updates and writes all its parental blocks. Therefore, any corruption to these blocks will become permanent after the flush caused by the access time update."
ZFS is not recommended for non-ECC RAM. RAM errors can get propagated to disk by application read operations, not just writes. http://research.cs.wisc.edu/ad...
Slashdot Mirror
The US Marine Corp used horses during the Korean War. Famously:
"She served in numerous combat actions during the Korean War, carrying supplies and ammunition, and was also used to evacuate wounded. Learning each supply route after only a couple of trips, she often traveled to deliver supplies to the troops on her own, without benefit of a handler. The highlight of her nine-month military career came in late March 1953 during the Battle for Outpost Vegas when, in a single day, she made 51 solo trips to resupply multiple front line units. She was wounded in combat twice."
https://en.wikipedia.org/wiki/...
The Marines still train to use pack animals in mountain warfare. I believe the US Army also has some training in the use of pack animals. I'm not sure if it is just for Special Forces or if "regular" Mountain Warfare units also have the training/capability.
Just to illustrate: "You say that like an Air Force Puke, I suppose you also think the F35 can replace the A10 too!" My guess, is he doesn't really know the difference between A and F aircraft, or else he wouldn't have conflated the roles like he did.
Actually the F-35 is considered a multimission aircraft, both a fighter and an attack aircraft. Hence it being referred to as the Joint Strike Fighter, "Strike" as in ground attack. And the Pentagon in fact says the F-35 is the "replacement" for the A-10, that the F-35 will provide close air support. And of course this is a controversial plan.
The A-10 is an anomaly, as is the original F-16, both being single purpose aircraft, and both originating from rogue teams working outside the normal Pentagon procurement process. The original F-16 could drop bombs but that capability was integrated with minimal compromises to air-to-air capability. Unlike a more traditional multimission aircraft where the tradeoffs are more dramatic. Later model F-16s moved a bit more toward multimission. The F-15 and F-14 were multimission from day one, as was the F-4 before them. Only the F/A-18 is truthful in its naming, "advertising" its multimission nature. Basically for decades if it originated from the Pentagon it was multimission.
Oh, and of course there is the F-117 which is in no way a fighter, pure ground attack. But was labeled as a fighter because they wanted to recruit fighter pilots for the program and to mislead foreign intelligence services.
In other words the use of "F" and "A" is very complicated.
Don't get stuck on today's missile designs. I expect the future AI enabled missiles to be larger, have larger control surfaces and a variable output engine so that they can go fast or maneuver as needed. In short they will be somewhat like a miniature aircraft. Think something "closer" to a cruise missile in look than a sparrow or sidewinder. Hmm ... those are more of yesterday's designs rather than todays but hopefully the point comes across.
Former Cav,Scout here.
My Dad too, that's why I'm familiar with the local Guard unit's history.
Every GOOD scout knows his greatest and most effective weapon... the RADIO.
Same as in my Dad's day many decades ago. That's why I wrote "A role not unlike that 1930s horse cavalry role, eyes, not direct combat." The radio is an indirect sort of combat. Basically, in my Dad's day and probably yours if the scout is shooting the scout is probably "seen", and you've described the problem with being seen.
Even though I detest Apple, they have prior art by about 8 years.
The Apple Newton
https://en.wikipedia.org/wiki/...
And lets not forget the tablets used in 2001: A Space Odyssey, a Stanley Kubrick film from 1968.
The big advantage of a pilot over a drone is that you can't jam or spoof a pilot. Same reason that wire-guided ATGMs remain a major player in modern battlefields.
The missiles I am suggesting will be human targeted and launched but once launched they will be self guided. Their sensors should be more capable than a human eyeball v1.0. Which is sort of what they currently do, but with better sensors and a more variable engine allowing for greater maneuverability, and an on-board AI to employ these, it will be a game changer. A fast stealthy aircraft is still useful as a launch platform, like a B-52 is today with cruise missiles, but the days of high maneuverability manned flight are drawing to a close. Again, in an air-to-air sense, for recon the current "fighter" capabilities will probably persist a while longer.
Electronic gunsights adjusting for target range and speed, telling the pilot when/where to shoot, date back to the 1950s and the Korean war. Not as fancy as a modern, or even somewhat modern hud, but the essentials are old tech.
If you can't see the difference between a poorly-maneuvering missile announcing its location from miles away and a fast jet actively engaging the target from within AAM range, your last sentence might apply to you more than me.
If you think missiles maneuver poorly compared to fighters, or are more detectable, you are indulging in the 420 too much. The only advantage fighters have over missiles are range. However even 1970s cruise missile technology has medium range low level terrain masking flight capabilities. They can be launched by bombers, tomorrow smaller versions will be launched by fighters. The AI will go into the missiles not the fighters. The fighters will become more of a launch platform for such missiles, like the B-52 launching cruise missiles today. Range and stealth will remain important, but high G maneuvering will not.
Missiles only seem low maneuverable, sometimes fighters avoiding one, because of their high speed. This high speed is needed to overtake a fleeing fighter. The high speed translates to a higher turning radius. Give missiles a propulsion system that is more variable and fighters will have little chance to outmaneuver them. Don't let the limitations of current technology confuse you.
As you say the Norden allowed more accurate bombing from high altitude but it was never anything near precise, even by WW2 standards. In WW2 when you wanted precision, in the sense that you hit a particular building, you sent in fighters or medium bombers (B-25, B-26) at low altitude. Heavy bombers (B-17, B-24) at altitude only hit a particular building through large numbers of bombs and statistical averages even with the Norden. The heavies were also more precise at low altitude but more vulnerable than the mediums. By the end of the war the heavies were found to be most effective at incendiaries where precision was not required.
That said, the Norden was an absolutely amazing piece of technology for its day.
The F35 and the A10, while both are "fighter" jets, have completely different roles. The F35 is for establishing air dominance so that the A10 can ground n pound.
Hopefully you are using "fighter" in quotes to really mean combat aircraft. But no, the A-10 is not a fighter. Its a ground attack aircraft. It carries air-to-air missiles as a last resort defense. Sort of like a member of a tank crew carrying a pistol. And like the tank crew member employing a pistol, if the A-10 pilot is employing air-to-air missiles something has gone terribly terribly wrong.
A lot of people overinterpret the lessons of the Korean war where missiles were overstressed versus the technology of the time... and have taken it as some universal lesson which will apply forever into the future, that close-range dogfighting will always be the most critical aspect of aircraft design.
Vietnam not Korea. Personally I expect the AI to go into the missiles not the aircraft. Fighters becoming a romantic anachronism, like horse cavalry. And like horse cavalry they will last longer than people expect. My local National Guard unit is cavalry, reconnaissance, and had horse as late as the 1930s. In certain terrain guys sneaking around on horse was still more effective than vehicles. They were just the eyes for armored formations and not expect to fight themselves. Sort of like modern Long Range Reconnaissance Patrols, if you are firing your weapons something has gone terribly wrong. Note some US Special Forces briefly operated as cavalry in Afghanistan. I believe the US Marines sometimes use dirt bikes. Recon may also be a role for repurposed fighters. Actually it has been such a role, removing guns an armor and adding cameras. Sometime there are gaps with satellite and drone coverage and a fast mover flying low and masking its approach with terrain fills that gap. A role not unlike that 1930s horse cavalry role, eyes, not direct combat.
Until they put that AI in a real plane, I won't believe the test.
The AI won't go into the plane, it will go into the missiles.
I've yet to see any form of public transit without windows. People wouldn't ride it.
Windows will become virtual. Your display can show you a movie, news, stock charts, etc or the outside world. And to be honest, imho, if the windows don't open to let in some air then virtual might not be that bad. Well, assuming, you are only a passenger and are not expected to take control of the vehicle at some point.
I know nothing about Chinese patent law, other than that WTO membership requires China to have one. But in the United States, the switch to "first to file" did not abridge the novelty requirement. Anything published, even other than as a patent application, before a given patent application is filed is considered prior art and therefore "filed".
No one said prior art no longer exists. The point is that Baili shipping first is not necessarily meaningful. If Apple filed before Baili shipped, and or course before Baili filed, then Apple "wins". Apple shipping second is not necessarily a problem as long as Apple filed early enough.
I think the patent system in China and in the US (now) is First to File. Discovering first or shipping first does not matter. So all that matters is the dates of the patent applications.
Strangely Intel supports ECC on the Skylake Celeron now.
Small and home office servers, appliances like NAS Raid boxes, etc. It would work great for a non-enterprise ZFS box.
View All shows 2011-2016 for E7, 2012-2016 for E5, but only 2011-2012 for E3. Apparently the E3 View All page is broken, apologies. The normal E3 landing page has an incomplete list of processors. It seems only the Compare Product Families page gives a complete list.
I did a BYO PC recently with an i7 gen 6. The comparable E3 v5 seem to either run hotter when the price is comparable or better, or are much more expensive when the they run with the same thermal output. Both situations would seem troublesome for Apple.
More cores?
Sorry, I was looking at E7 Xeons, 8-10 cores for the 4-5 year old contemporaries of the E3s. Up to 24 cores for the current E7s.
When did I mention 4-5 year old Xeons?
The E3 series, the most recent versions released in mid 2012. Use the link you provided and select View All E3. Notice the 2011-12 launch dates.
For you it may be a low risk. For Apple its not. Apple will be shipping millions of machines.
And these machines are already vulnerable just to single bit errors anywhere both in the IO path and in memory. The repair-of-death you describe involves multiple errors ...
And that is how airplanes occasionally crash. Its usually not one flaw or problem, its multiple problems/flaws occurring at the same time. If you are an individual wishing to fly this is of little concern, the odds are with you. If you are an aircraft manufacturer it is of concern since your aircraft will conduct millions of flights and there are hostile lawyers eyeing your deep pockets.
Apple was in a similar situation, but with deeper pockets to attract class action lawyers.
Stop approaching this from the perspective that ZFS is flawed. Rather approach this from the perspective that ZFS assumes memory can be trusted
... so does every other filesystem. I'll quote another bit of that paper you like:
There is quite a difference between corrupting the inode info / timestamp info and corrupting the **contents of a file**, its user data. That is what is unique about ZFS. File data being **read** is at risk due to automatic repairs of **user data**.
Intel is more likely to support ECC in lower end CPUs (ex i3) than in mid to higher end CPUS (ex i5, i7)
i7-class Xeons (E3-XXXX) support ECC and are usually priced basically identically to their i7 cousins.
Yes, server grade CPUs support server grade RAM. And judging from Intel's data sheets the current generation Xeons are slower (clock rate, more cores though) and generate more heat. Notice Apple's consumer products are designed for more moderate thermal conditions. And the 4-5 year old Xeons you mention seem to be slightly more expensive that the new i7s in Intel's price lists.
Another difficulty for a consumer oriented company like Apple, making using ECC not really an option for them
I'm sure Apple are more than capable of pushing for it if they considered it a priority. They have the purchasing power, they have the margins, they have the PR to make people wet themselves over the benefits if they so choose.
Intel used to support ECC in some i5 and i7, if I'm reading the 6th gen data sheets correctly no i5 or i7 supports ECC. I severely doubt Apple can change Intel's direction. Plus, no, consumers won't care so why would Apple even bother? ZFS was interesting when they offered servers, they no longer do so, the interest is gone, its not their mission to advocate for ZFS.
So use ECC RAM. By an amazing coincidence, the company announcing this software also just happens to also manufacture hardware.
ECC requires a compatible motherboard, CPU and RAM. Strangely higher end Intel CPUs, i5 and i7, tend not to support ECC.
The bit flips required to make such a process actually damage your data seems quite convoluted - it'd have to be multiple errors in different locations happening at just the right times ...
For you it may be a low risk. For Apple its not. Apple will be shipping millions of machines. Plus given prior knowledge of ZFS' unique failure mode they will be vulnerable in court, a hostile class action lawyer arguing they increased their customer's vulnerability through their choice of filesystem and type of RAM.
Is ZFS really significantly more vulnerable to this by comparison, or is it just that ZFS won't defend you against it?
More vulnerable. The important distinction is that ZFS can corrupt on-disk data when applications are performing read operations. That is what differentiates ZFS from other filesystems.
Stop approaching this from the perspective that ZFS is flawed. Rather approach this from the perspective that ZFS assumes memory can be trusted and should be used with system equipped with ECC RAM. Every piece of software has system requirements. ZFS just has an extra one. Given ZFS' capabilities it an excellent tradeoff, but its not a tradeoff available to consumer hardware.
I'm going the BYO route for my SOHO server, as I have been doing for PCs for decades. Motherboard, CPU and RAM all supporting ECC. For those reading along you need all three, and strangely Intel is more likely to support ECC in lower end CPUs (ex i3) than in mid to higher end CPUS (ex i5, i7). Another difficulty for a consumer oriented company like Apple, making using ECC not really an option for them. ZFS is just not a good fit for Apple now that they no longer sell server grade equipment, hence their losing interest.
Nothing you're saying there is unique to ZFS.
No, ZFS is more vulnerable. All filesystems can corrupt the disk by writing bad data, however ZFS can corrupt data on disk when an application only reads the disk. When checksums fail ZFS will assume the problem is on disk and attempt to "repair" the data on disk. This automatic repair is a great feature, when your RAM can be trusted.
One example from the cited paper: "Dirtying blocks due to updating file access time increases the possibility of making corruptions permanent. By default, access time updates are enabled in ZFS; therefore, a read-only workload will update the access time of any file accessed. Consequently, when the structure containing the access time (znode) goes inactive (or when there is another workload that updates the znode), ZFS writes the block holding the znode to disk and updates and writes all its parental blocks. Therefore, any corruption to these blocks will become permanent after the flush caused by the access time update."
I am not saying it is a new risk in ZFS, merely that it is something understood about ZFS for a long time. And given that Apple is making consumer devices and does not want to use ECC, ZFS is not a good fit and that is most likely why Apple passed on it. If Apple were still making servers then ZFS may very well be an option on those.
You said "ZFS needs ECC RAM". Implying that ZFS is special is this regard. The link however contains the following:
ZFS needs ECC as much as every other file system.
No, ZFS is more vulnerable. All filesystems can corrupt the disk by writing bad data, however ZFS can corrupt data on disk when an application only reads the disk. When checksums fail ZFS will assume the problem is on disk and attempt to "repair" the data on disk. This automatic repair is a great feature, when your RAM can be trusted.
Ex: "Dirtying blocks due to updating file access time increases the possibility of making corruptions permanent. By default, access time updates are en- abled in ZFS; therefore, a read-only workload will update the access time of any file accessed. Consequently, when the structure containing the access time (znode) goes inactive (or when there is another workload that updates the znode), ZFS writes the block holding the znode to disk and updates and writes all its parental blocks. Therefore, any corruption to these blocks will become permanent after the flush caused by the access time update."
ZFS is not recommended for non-ECC RAM. RAM errors can get propagated to disk by application read operations, not just writes.
http://research.cs.wisc.edu/ad...