It's probably more accurate to say that the version in Solaris is a fork of an older version. Most of the ZFS developers left Oracle quite early on after they bought Sun and most of the rest left when Oracle decided to stop releasing CDDL versions of Solaris. The version that ended up in OpenZFS has been actively developed by the same people who created ZFS for the last 10 years. The version that Oracle owns has had a few incremental changes. This also means that it would be difficult for Oracle to GPL ZFS in a useful way: the version that's had all of the work done on it for the last decade contains a load of CDDL code that Oracle doesn't own.
I use ZFS on a NAS with a bunch of drives, but I also use it on a hosted VM with under 1GB of RAM on a single (virtual) drive and a few local VMs. The benefits that I'm apparently not getting include:
It's trivial to add more storage. If I want to expand a VM, I attach a new virtual disk and simply expand my storage onto it.
It's trivial to back up - I can snapshot all of my ZFS filesystems and use zfs send / zfs receive to send incremental snapshots of them to another system (where I can reconstruct all of the snapshots and have the ability to look at earlier configs, or I can simply store the latest one).
Block-level checksums mean that I can detect errors early, even if I can't recover from them unless I've set copies=2 or more on a filesystem.
I can snapshot a filesystem before I make big config changes, so if I make a mistake I have an easy undo function.
I get transparent LZ4 compression, which saves on storage space and bandwidth.
I can run jails inside my VMs that use CoW copies of a single system install, so have very low space overhead
Gold is a good conductor and doesn't tarnish, and so has value independent of speculation. The current value of gold owes a lot more to speculation than to these properties, but if everyone decided tomorrow that gold was a poor store of wealth then its value would be non-zero. The same does not hold for Bitcoin.
Yes, and they will lose if they are slower than you. It's a zero-sum game, but as long as there are people willing to play then some will win and some will lose.
Any kind of teleporter that works by scanning an object and rebuilding it hits problems with the uncertainty principle, so this one makes sense: you can see with current physics that there is a problem that needs solving, you wouldn't expect to understand the solution that's based on technology from 200 years in the future.
Two more: Warp drive
The science behind an Alcubierre drive is pretty well understood, though it's not clear whether the exotic matter required to build one can actually exist in this universe. This is magic technology, but not magic science. Again, plausible with a couple of hundred years of technological advance.
Two more: Deflector dish
Not sure I see the issue with this one. We have dishes for transmitting all kinds of RF signals now, using them to transmit whatever is used to generate deflector shields doesn't seem to unlikely.
Two more: Inertial dampeners
This one is actually explained a few times in-show. At sub-light speeds (i.e. when you're not taking an inertial frame of reference with you), acceleration is a problem. If you have artificial gravity, then having it automatically adjust to compensate for acceleration makes sense.
Two more: Artificial gravity (not even mentioned, simply handwaved)
Again, assuming that 200 years more scientific and technological advancement will allow manipulation of gravity isn't too far fetched, especially given the ability to create the kinds of exotic matter required for a warp drive.
Single words: Transporter/Replicator, Shields, Phasers
Replicator is a natural offshoot of the transporter (both of which are explained in-show a few times). Shields and phasors are hand waved. The TOS pilot used lasers and it's later implied that phasors are a development along the same lines. About 20 years ago, there was a paper published proposing that you could achieve the stun setting of a phasor by using a laser to ionise the air and then send an electrical charge, effectively a wireless Taser. It's hard to do currently (the required laser power would burn the victim quite badly), but again it's plausible that this could be addressed with 200 years of advancement. Shields, again, don't seem too much of a stretch with a technology that has the ability to control the electromagnetic and gravitational forces - assuming technology that can do that, there are lots of ways you can imagine producing shields.
And it's just as much fiat as most other currencies
No it isn't. Fiat isn't just a word Libertarians throw around, it has an actual meaning. Literally, the value of a fiat currency is imposed by decree. More practically, the requirement for a government (and, usually, other people) to accept the currency is imposed by law. The IRS, for example, is required to accept payments for taxes in USD. This guarantees that there are always going to be buyers for USD as long as there are taxpayers in the US: even if all of your other income and expenditures are in Euros, Bitcoin, or Disney Tokens, you need to be able to produce a sum of US dollars to pay taxes if you are a US taxpayer.
This is in contrast to commodity-backed currencies. The earliest recorded currency was backed by grain (I forget which kind and I'm too lazy to look it up now). The city granary guaranteed that it would pay you a fixed amount of grain in exchange for the token. The currency had value because everyone needs to eat and so there were always people who wanted to be able to go to the granary. Later commodity-backed currencies were backed by precious metals, on the basis that there is an endless supply of idiots who want a shiny thing.
The move away from commodity-backed currencies was due to many factors, but one was that currencies behave as commodities in their own right and pegging the value of one commodity to another is awkward. If there's more demand for Pounds Sterling than there is for sterling silver, then the value of the pound goes up, but the linking of value means that the value of silver goes up artificially and the only way to fix it is to find a pile of silver somewhere.
The problem with bitcoin is that it has neither value from people who want to exchange it for something at a fixed rate, nor value from people who have a legal requirement to pay for certain things using it. It has value solely from the fact that people use it for speculation. As soon as the speculators move on to the next shiny unregulated high-volatility commodity, the value of Bitcoin goes away.
If drug dealers still accept Bitcoin, they are likely to stop soon. Bitcoin is a public ledger of all transactions. If someone is identified as a drug dealer then it's trivial to backtrack all of their Bitcoin receipts and find all of the people that they've sold things to, and all of the people that they've bought things from.
The value of Bitcoin for speculators is its volatility. A commodity that fluctuates in value by 10-20% per day means that you can make huge returns by selling while it's high and buying while it's low. Lots of speculators doing this increase the volatility: when the value is low, lots of people buy and the value goes up. This will likely continue until something else looks more attractive, at which point the people holding Bitcoin will all try to sell them (slowly, if they're sensible) and the value will crash.
I have a Moto G, first generation, which I bought a few months after it was launched. The USB port is still working fine, and it's plugged in overnight for charging every day. Maybe you're inserting the cable too hard? Micro USB was designed to be more resilient than the older mini USB, and seems pretty solid to me.
China became a nuclear power in the '60s, but I don't know when they developed sufficient launch capability to threaten the US. There are a number of references to China's nuclear capability in popular culture from the late '60s, which is why I'm surprised that they weren't at least the target for a couple of silos. That and the desire for some pork-barrel spending for the west coast.
That sounds a lot more sensible: Windows NT has had ACLs (much richer than the default UNIX model and similarly expressive to NFSv4 / POSIX ACLs) since day one, but the ACLs have been per user, not per (user, program) pair. The NT kernel supports this kind of ACL policy, but it's never been exposed via the UI (Chromium uses it for sandboxing, constraining different binaries to different parts of the FS).
It's very useful if it's paired with a sensible default policy and a sensible UI. You can implement the same thing with the TrustedBSD MAC framework or SELinux, and macOS / iOS implement their sandboxing policies in exactly this way. macOS, in particular, provides a 'powerbox' model, where the standard open and save dialogs are owned by the system and implicitly grant the application permissions to the files / directories that the user selects as part of a dynamic policy. This means that well-behaved applications never need to ask for explicit privilege elevation. The problem is, well-behaved applications are generally not the ones that you most want to sandbox...
I don't think that's the market that they're going for, because if that's what you want then just look at the installed base for LineageOS and pick one of the ones with a lot of users (I'm still using a first-gen Moto G) - something like one of the onePlus line is well supported by LineageOS and has decent hardware.
The implicit assumption in your reasoning is that China wants to stop this kind of confrontation. Weakening the US economically and severely damaging Japan could both be seen as being in China's interest, so it's not entirely clear whether they'd actually want to step in - especially if a conflict would result in the rest of the world turning against the USA politically. And if they wished to escalate, they could easily promise to defend NK against US missile and air attack, only for the support to never quite materialise at the last minute.
It's an Android phone, that looks and feels like an Android phone. What makes it unique? Reading their web site, I'm not seeing anything compelling. Anyone bought one and found something unique in it?
Really? I'd have thought that, during the cold war, China would have been considered a big enough threat to warrant a few missiles pointed their way, and anything that can hit China from the US should be able to hit North Korea.
That's the reason you want bombers and subs in a war against a country like Russia or China - because their first-strike capability may be enough to take out any ground-based launch capability. It's a vital part of MAD. The worst thing that you can possibly do if you are a nuclear power is have a solid first-strike capability and no second strike. The winning strategy for an opponent if you do that is to launch a first strike, because they know that the only way for you to win is to attack first, and it's quite troubling to have an opponent in that situation.
This is, by the way, a big problem for North Korea: until they finish their submarine launch platforms, they have no real second-strike capability. They might be able to load a nuke into one of their existing subs and detonate it in a US harbour, but they aren't in a position to say 'attack us and New York is toast'.
North Korea, unlike the USSR, has nowhere near the capability to destroy the US military in a first strike. Even if they managed to launch all of their nuclear weapons at their desired targets with no interception, they couldn't destroy the US first-strike capability, so there is no need for the US to bring its second-strike capability to readiness other than for dick measuring.
Define 'destroy'. If North Korea successfully destroyed New York and San Francisco, how much of a hit do you think the US economy would take? If they managed to sail submarines into the five largest trading ports and detonate nuclear weapons, first hitting the ports with a tsunami and then covering them in radioactive water and making them unsafe to use for a long time, what do you think that would do to the US? Even ignoring the deaths, can you imagine the cost of trying to rearrange logistics in the US to avoid imports and exports from, say, New Jersey of Houston?
There is no winning a nuclear war, ever. Radiation from Japan made it all the way to the west coast. The bombs of those days might as well have been a stick of dynamite compared to what we could release today.
Both are mostly true, but the implication doesn't necessarily follow. One of the reasons that modern nuclear weapons are higher yield is that they are more efficient. 'Radiation' didn't travel from Japan to the West Coast of the USA, because radiation only travels in straight lines and is blocked quite well by the curve of the planet. Radioactive fallout travelled that far. Radioactive fallout is the leftover radioactive material that is not consumed in the nuclear reaction and is dispersed by the explosion. In other words, it's a waste byproduct of inefficient nuclear weapons. The more efficient the weapon, the less radioactive material is left after the explosion to become fallout.
There's no such thing as a completely clean nuke (in theory there could be, but nuclear weapons are trivial to build in theory, it's only the engineering that's hard), but a modern weapon can have a much higher yield for the same amount of fallout.
ASLR in userspace increases work factor, but attacks such as BROP and successors can bypass it. It can also suffer from various weakness - for example the StageFright vulnerability on Android was made worse by the fact that, on 32-bit systems, jemalloc allocated in large chunks (and didn't randomise within a chunk) and so you ended up with 8 bits of entropy, and the automatic restart meant that, on average, you could guess (and get root privilege arbitrary code execution) in 128 attempts.
KASLR, in contrast, is entirely snake oil. Kernel interfaces were never designed to avoid leaking kernel addresses to unprivileged code, because kernel security doesn't rely on addresses being secret, it relies on kernel addresses being unusable from userspace. The kernel-userspace interfaces (system calls and ioctls) provide a large number of ways of finding kernel addresses from userspace. If your threat model is protecting the OS from a malicious device via DMA or a malicious hypervisor, then it's even weaker.
More RAM stops helping when your entire working set fits in RAM. If you have enough RAM, then when you're machine has been on for a while you write to disk but never read things back. For most people, that amount is still more than 32GB. Even with SSDs, there's a noticeable performance difference between getting data from RAM and from SSD.
That said, I agree that SSD vs spinning rust is a far more noticeable performance win for most workloads.
A lot of modern smartphones and tablets have HDMI output, so you can carry them around in your pocket and plug them in to a big screen plus a bluetooth keyboard and mouse. Of course, then you're stuck with a somewhat underpowered device with a UI that's designed for a 5" screen and doesn't scale well to a larger display.
That said, I'd love for someone to resurrect the Samsung research project from about 10 years ago that did partial migration with Xen. They had a demo where they ran a VM on a phone, then plugged it into a big computer and used the OS hotplug facilities to make it think that it now had more cores, more RAM, and some extra peripherals, in a NUMA arrangement: memory pages were automatically faulted across between the machines by the hypervisor and if the scheduler moved a process to the fast CPUs it would eventually move over. The only real problem was that everything broke horribly if you unplugged the phone before migrating everything back, but that could be solved by providing a dock that doesn't release the phone until it's either powered off or everything is migrated back. This arrangement let you use a small mobile device for everyday computing, but move large workloads transparently to more powerful compute resources when you needed them.
We typically have a 3-year rolling upgrade program, but my work laptop is now 4 years old. The newer Intel CPUs are only about 20-30% faster and until they start supporting LPDDR4, I'm stuck with a choice of 16GB of RAM (the same as my current machine) or a crappy battery life (10W+ idle power consumption for 32GB of DDR4). GPUs are quite a bit faster, but I don't use the GPU much for work.
There's an easy way to avoid leaking a load of personal information when you're hacked: Don't keep a load of personal information on your servers. If regulation moved the default from 'let's keep everything, it may be useful for something eventually' to 'don't keep anything unless you can demonstrate a really strong business case that requires keeping it and outweighs the cost of insurance for the potential liability if it's leaked' then that would be a huge improvement.
Slashdot Mirror
It's probably more accurate to say that the version in Solaris is a fork of an older version. Most of the ZFS developers left Oracle quite early on after they bought Sun and most of the rest left when Oracle decided to stop releasing CDDL versions of Solaris. The version that ended up in OpenZFS has been actively developed by the same people who created ZFS for the last 10 years. The version that Oracle owns has had a few incremental changes. This also means that it would be difficult for Oracle to GPL ZFS in a useful way: the version that's had all of the work done on it for the last decade contains a load of CDDL code that Oracle doesn't own.
I use ZFS on a NAS with a bunch of drives, but I also use it on a hosted VM with under 1GB of RAM on a single (virtual) drive and a few local VMs. The benefits that I'm apparently not getting include:
Gold is a good conductor and doesn't tarnish, and so has value independent of speculation. The current value of gold owes a lot more to speculation than to these properties, but if everyone decided tomorrow that gold was a poor store of wealth then its value would be non-zero. The same does not hold for Bitcoin.
Yes, and they will lose if they are slower than you. It's a zero-sum game, but as long as there are people willing to play then some will win and some will lose.
Two words: Heinsenberg compensators.
Any kind of teleporter that works by scanning an object and rebuilding it hits problems with the uncertainty principle, so this one makes sense: you can see with current physics that there is a problem that needs solving, you wouldn't expect to understand the solution that's based on technology from 200 years in the future.
Two more: Warp drive
The science behind an Alcubierre drive is pretty well understood, though it's not clear whether the exotic matter required to build one can actually exist in this universe. This is magic technology, but not magic science. Again, plausible with a couple of hundred years of technological advance.
Two more: Deflector dish
Not sure I see the issue with this one. We have dishes for transmitting all kinds of RF signals now, using them to transmit whatever is used to generate deflector shields doesn't seem to unlikely.
Two more: Inertial dampeners
This one is actually explained a few times in-show. At sub-light speeds (i.e. when you're not taking an inertial frame of reference with you), acceleration is a problem. If you have artificial gravity, then having it automatically adjust to compensate for acceleration makes sense.
Two more: Artificial gravity (not even mentioned, simply handwaved)
Again, assuming that 200 years more scientific and technological advancement will allow manipulation of gravity isn't too far fetched, especially given the ability to create the kinds of exotic matter required for a warp drive.
Single words: Transporter/Replicator, Shields, Phasers
Replicator is a natural offshoot of the transporter (both of which are explained in-show a few times). Shields and phasors are hand waved. The TOS pilot used lasers and it's later implied that phasors are a development along the same lines. About 20 years ago, there was a paper published proposing that you could achieve the stun setting of a phasor by using a laser to ionise the air and then send an electrical charge, effectively a wireless Taser. It's hard to do currently (the required laser power would burn the victim quite badly), but again it's plausible that this could be addressed with 200 years of advancement. Shields, again, don't seem too much of a stretch with a technology that has the ability to control the electromagnetic and gravitational forces - assuming technology that can do that, there are lots of ways you can imagine producing shields.
And it's just as much fiat as most other currencies
No it isn't. Fiat isn't just a word Libertarians throw around, it has an actual meaning. Literally, the value of a fiat currency is imposed by decree. More practically, the requirement for a government (and, usually, other people) to accept the currency is imposed by law. The IRS, for example, is required to accept payments for taxes in USD. This guarantees that there are always going to be buyers for USD as long as there are taxpayers in the US: even if all of your other income and expenditures are in Euros, Bitcoin, or Disney Tokens, you need to be able to produce a sum of US dollars to pay taxes if you are a US taxpayer.
This is in contrast to commodity-backed currencies. The earliest recorded currency was backed by grain (I forget which kind and I'm too lazy to look it up now). The city granary guaranteed that it would pay you a fixed amount of grain in exchange for the token. The currency had value because everyone needs to eat and so there were always people who wanted to be able to go to the granary. Later commodity-backed currencies were backed by precious metals, on the basis that there is an endless supply of idiots who want a shiny thing.
The move away from commodity-backed currencies was due to many factors, but one was that currencies behave as commodities in their own right and pegging the value of one commodity to another is awkward. If there's more demand for Pounds Sterling than there is for sterling silver, then the value of the pound goes up, but the linking of value means that the value of silver goes up artificially and the only way to fix it is to find a pile of silver somewhere.
The problem with bitcoin is that it has neither value from people who want to exchange it for something at a fixed rate, nor value from people who have a legal requirement to pay for certain things using it. It has value solely from the fact that people use it for speculation. As soon as the speculators move on to the next shiny unregulated high-volatility commodity, the value of Bitcoin goes away.
If drug dealers still accept Bitcoin, they are likely to stop soon. Bitcoin is a public ledger of all transactions. If someone is identified as a drug dealer then it's trivial to backtrack all of their Bitcoin receipts and find all of the people that they've sold things to, and all of the people that they've bought things from.
The value of Bitcoin for speculators is its volatility. A commodity that fluctuates in value by 10-20% per day means that you can make huge returns by selling while it's high and buying while it's low. Lots of speculators doing this increase the volatility: when the value is low, lots of people buy and the value goes up. This will likely continue until something else looks more attractive, at which point the people holding Bitcoin will all try to sell them (slowly, if they're sensible) and the value will crash.
I have a Moto G, first generation, which I bought a few months after it was launched. The USB port is still working fine, and it's plugged in overnight for charging every day. Maybe you're inserting the cable too hard? Micro USB was designed to be more resilient than the older mini USB, and seems pretty solid to me.
Thanks. I looked it up and you're entirely correct.
China became a nuclear power in the '60s, but I don't know when they developed sufficient launch capability to threaten the US. There are a number of references to China's nuclear capability in popular culture from the late '60s, which is why I'm surprised that they weren't at least the target for a couple of silos. That and the desire for some pork-barrel spending for the west coast.
It's very useful if it's paired with a sensible default policy and a sensible UI. You can implement the same thing with the TrustedBSD MAC framework or SELinux, and macOS / iOS implement their sandboxing policies in exactly this way. macOS, in particular, provides a 'powerbox' model, where the standard open and save dialogs are owned by the system and implicitly grant the application permissions to the files / directories that the user selects as part of a dynamic policy. This means that well-behaved applications never need to ask for explicit privilege elevation. The problem is, well-behaved applications are generally not the ones that you most want to sandbox...
I don't think that's the market that they're going for, because if that's what you want then just look at the installed base for LineageOS and pick one of the ones with a lot of users (I'm still using a first-gen Moto G) - something like one of the onePlus line is well supported by LineageOS and has decent hardware.
The implicit assumption in your reasoning is that China wants to stop this kind of confrontation. Weakening the US economically and severely damaging Japan could both be seen as being in China's interest, so it's not entirely clear whether they'd actually want to step in - especially if a conflict would result in the rest of the world turning against the USA politically. And if they wished to escalate, they could easily promise to defend NK against US missile and air attack, only for the support to never quite materialise at the last minute.
It's an Android phone, that looks and feels like an Android phone. What makes it unique? Reading their web site, I'm not seeing anything compelling. Anyone bought one and found something unique in it?
Really? I'd have thought that, during the cold war, China would have been considered a big enough threat to warrant a few missiles pointed their way, and anything that can hit China from the US should be able to hit North Korea.
This is, by the way, a big problem for North Korea: until they finish their submarine launch platforms, they have no real second-strike capability. They might be able to load a nuke into one of their existing subs and detonate it in a US harbour, but they aren't in a position to say 'attack us and New York is toast'.
North Korea, unlike the USSR, has nowhere near the capability to destroy the US military in a first strike. Even if they managed to launch all of their nuclear weapons at their desired targets with no interception, they couldn't destroy the US first-strike capability, so there is no need for the US to bring its second-strike capability to readiness other than for dick measuring.
Define 'destroy'. If North Korea successfully destroyed New York and San Francisco, how much of a hit do you think the US economy would take? If they managed to sail submarines into the five largest trading ports and detonate nuclear weapons, first hitting the ports with a tsunami and then covering them in radioactive water and making them unsafe to use for a long time, what do you think that would do to the US? Even ignoring the deaths, can you imagine the cost of trying to rearrange logistics in the US to avoid imports and exports from, say, New Jersey of Houston?
There is no winning a nuclear war, ever. Radiation from Japan made it all the way to the west coast. The bombs of those days might as well have been a stick of dynamite compared to what we could release today.
Both are mostly true, but the implication doesn't necessarily follow. One of the reasons that modern nuclear weapons are higher yield is that they are more efficient. 'Radiation' didn't travel from Japan to the West Coast of the USA, because radiation only travels in straight lines and is blocked quite well by the curve of the planet. Radioactive fallout travelled that far. Radioactive fallout is the leftover radioactive material that is not consumed in the nuclear reaction and is dispersed by the explosion. In other words, it's a waste byproduct of inefficient nuclear weapons. The more efficient the weapon, the less radioactive material is left after the explosion to become fallout.
There's no such thing as a completely clean nuke (in theory there could be, but nuclear weapons are trivial to build in theory, it's only the engineering that's hard), but a modern weapon can have a much higher yield for the same amount of fallout.
ASLR in userspace increases work factor, but attacks such as BROP and successors can bypass it. It can also suffer from various weakness - for example the StageFright vulnerability on Android was made worse by the fact that, on 32-bit systems, jemalloc allocated in large chunks (and didn't randomise within a chunk) and so you ended up with 8 bits of entropy, and the automatic restart meant that, on average, you could guess (and get root privilege arbitrary code execution) in 128 attempts.
KASLR, in contrast, is entirely snake oil. Kernel interfaces were never designed to avoid leaking kernel addresses to unprivileged code, because kernel security doesn't rely on addresses being secret, it relies on kernel addresses being unusable from userspace. The kernel-userspace interfaces (system calls and ioctls) provide a large number of ways of finding kernel addresses from userspace. If your threat model is protecting the OS from a malicious device via DMA or a malicious hypervisor, then it's even weaker.
More RAM stops helping when your entire working set fits in RAM. If you have enough RAM, then when you're machine has been on for a while you write to disk but never read things back. For most people, that amount is still more than 32GB. Even with SSDs, there's a noticeable performance difference between getting data from RAM and from SSD.
That said, I agree that SSD vs spinning rust is a far more noticeable performance win for most workloads.
A lot of modern smartphones and tablets have HDMI output, so you can carry them around in your pocket and plug them in to a big screen plus a bluetooth keyboard and mouse. Of course, then you're stuck with a somewhat underpowered device with a UI that's designed for a 5" screen and doesn't scale well to a larger display.
That said, I'd love for someone to resurrect the Samsung research project from about 10 years ago that did partial migration with Xen. They had a demo where they ran a VM on a phone, then plugged it into a big computer and used the OS hotplug facilities to make it think that it now had more cores, more RAM, and some extra peripherals, in a NUMA arrangement: memory pages were automatically faulted across between the machines by the hypervisor and if the scheduler moved a process to the fast CPUs it would eventually move over. The only real problem was that everything broke horribly if you unplugged the phone before migrating everything back, but that could be solved by providing a dock that doesn't release the phone until it's either powered off or everything is migrated back. This arrangement let you use a small mobile device for everyday computing, but move large workloads transparently to more powerful compute resources when you needed them.
We typically have a 3-year rolling upgrade program, but my work laptop is now 4 years old. The newer Intel CPUs are only about 20-30% faster and until they start supporting LPDDR4, I'm stuck with a choice of 16GB of RAM (the same as my current machine) or a crappy battery life (10W+ idle power consumption for 32GB of DDR4). GPUs are quite a bit faster, but I don't use the GPU much for work.
AMD's ARM strategy is... complicated. As far as I can tell, they are very enthusiastic in their desire to plan to have a plan.
There's an easy way to avoid leaking a load of personal information when you're hacked: Don't keep a load of personal information on your servers. If regulation moved the default from 'let's keep everything, it may be useful for something eventually' to 'don't keep anything unless you can demonstrate a really strong business case that requires keeping it and outweighs the cost of insurance for the potential liability if it's leaked' then that would be a huge improvement.