I know a guy who works for a company that does a lot of lab testing of its products. Many years ago they were outsourcing quite a bit of testing to a number of companies and getting cheap prices. However, they had all kinds of quality issues - the various labs were not consistent with each other and the bottom-line mentality was making it hard to use the data that was generated.
So, they consolidated all their labwork to a single contract facility (the one that consistently had the best quality). Costs went up, but it was still cheaper than in-house, and the volume led to a partnership relationship with the testing lab, which was eager to compete on quality instead of price. Quality and throughput were better than ever, and cost was still very reasonable.
Fast forward a few years, and now the big rage is outsourcing. Now they're trying to go back to getting 5 vendors bidding against each other to squeeze every nickel of cost out of the system. How long do you think it will be before bad decisions are made on the basis of bad data?
When you get an excellent plasma TV for $2000 instead of $3000 you just saved $1000. When you get a broken plasma TV for $1500 instead of $2000 you didn't save $500 - you wasted $1500.
Maybe the ivory tower folks might not understand this, but perhaps the average American would rather be hung in a harness for 5 hours than to spend 30 hours on a train, or spending an extra $5000 for an airline ticket?
I agree that takeoff/landing slots should be limited to the quantity that can be safely handled. Probably the easiest way to handle this is via auction for 90% of the slots, with a fairly high fee for the last 10% so that non-advance-planned flights can still use the airport. Then consumers can choose whether they want to pay an extra $50 to be able to sit down on their flight. What's wrong with that?
The ultra-rich can still fly their private jets in complete comfort, although they would need to pay into the ATC maintenance like everybody else. They would pay a disproportionate share since the fees would be based on the cost to service an aircraft, and it is a heck of a lot cheaper per passenger to takeoff/land an A380 than a Gulfstream.
And maybe we can actually get to fixing the actual problem and stop putting band-aids over it. I'm not quite sure why humans are so heavily involved in the ATC system. You'd probably be far more efficient with an automated system that calls on humans when it needs help, and which relays instructions to planes digitally instead of making humans call out instructions and then read them back for verification. There is no reason that 95% of the time the computer in ATC can't just talk to the computer flying the plane, with humans available for failure modes and in case a little personal touch is needed.
Somehow I doubt that will work with any frequency - you'd think newegg would wise up after the 3rd return with tool marks on it. If it did brick and mortar stores (which commonly have this policy) would be completely out of business.
If the item didn't have damage upon shipment, and it has a big bash in it when it is returned, then I'd claim for shipping insurance. If the insurer notices a trend in activity from a certain address then they're going to prosecute for insurance fraud.
There are lots of solutions to this problem. You don't need to stick it to your perfectly honest customers (the vast majority) just to get rid of a few abusers.
I agree - RMA for return for defective items should NOT have a restocking fee.
Sure - if the return is because "I don't like it" I'm fine with restocking fees. It helps a company like newegg keep costs down.
Also - returns should be parallel-ship - you ask for an RMA and the replacement item goes out immediately. If the return doesn't show up in x days charge the customer for a second item. None of that waiting around for mail to go back and forth.
And postage-paid returns are perfectly reasonable. It drives me nuts to have to pay to ship an item back if it is defective. I'm sorry - if a merchant gets defective junk they should tell their supplier to eat the costs - or threaten to drop the product. I'm Seagate would care if Newegg threatened to drop their product line. Sure, maybe it isn't cost-effective to test every item, but they can at least pay for the ones that leak through.
Most likely it is related to a motherboard that doesn't handle hard drive failures well - a dying hard drive was causing some flaky behavior. Even so, I can't see why that would cause LVM to cause data access beyond volume boundaries.
But I'll go ahead and echo your sentiments. I was running ext3+lvm+raid for a while without any issues at all. So are most people. However, I'll probably not do it again with data that I can't fully back up for a long time. Perhaps I'm just a lucky 1 out of 10,000. This is a DVR, so I'm not going to invest in what it takes to back up a few hundred GB of data offsite for some TV - nothing I lost wasn't expendible. It just was a real pain reinstalling everything.
Also - keep in mind that even enterprise grade hardware fails. The failure mode of the underlying software is still important.
I don't really see how this would be a distro issue. I was running versions of kernel+mdadm+lvm+ext3-tools that the upstream providers would have considered stable. There really isn't much else that should interact with the disks at that level.
I used fdisk to create partitions on the drives. Then I used mdadm to create RAID-5s on the partitions. Then I turned the/dev/md# raid devices into physical volumes and combined them into a volume group. Then I created logical volumes on top of that.
I can't see how the partitions couldn't have been valid. They start on one cylinder, and end on another. Not much more to it than that.
I didn't have any trouble with partitions that resided directly on md devices - such as boot/root. Only the logical volumes had problems. I also had a flaky hard drive at the time which was causing frequent crashes - once I figured out the cause the system was stable, but by then I had already hosed everything with an fsck. I figured that I couldn't go too wrong just running that on my video logical volume - after all that was all expendible. I didn't count on it corrupting stuff left and right on all my other logical volumes. With all the crashes the system did spend quite a bit of time in recovery mode - which might not have helped - a crash in recovery mode can't necessarily be cleaned up at the RAID level - so damage could leak into the LVM layer. Even so, the sector mapping used by LVM should only change when adding new logical volumes, and it should never be in a state that would cause one partition to overwrite another. I can't see how a crash should result in LVM allowing writes outside of logical volume boundaries.
Actually - the issue isn't that myth was single-threaded. The problem is that the filesystem itself totally blocks everything when a delete is in progress.
Myth didn't just move the delete into a separate thread. It also turned the delete into a piece-by-piece truncation of the file. This way the filesystem is never deleting more than a small amount of data at a time and other stuff can get in beteween the deletes.
Go ahead and try creating two separate processes - have one keep reading from a file (typical playback process), and then have an entirely different process delete a 20GB file while that is running. Watch the playback choke (assuming it isn't all cached in RAM).
The last time I checked it was also not possible to reshape a raidz. As in - I have 5 drives in a RAID, and now I want to have 6 drives in a raid. Then, I have 6 250GB drives in a RAID and I want to switch to 6 500GB drives in the RAID and get all the space without resorting to splitting the drives into 2x250GB partitions. I'm not talkinga bout adding another raidz to a pool containing a raidz.
Linux software RAID5 supports reshaping in this way (I think it even supports it online for the brave).
Don't get me wrong - I love the whole idea of ZFS and I think that copy-on-write will be the future of filesystems. However, the implementation isn't as mature as other stuff out there quite yet. I'm sure we'll get there eventually - there is no theoretical reason you couldn't reshape a raidz at least offline.
I'll vote for the "doesn't think it is important" option. That was the one thing that always annoyed me about MySQL. Rather than apologize for taking a while to support transactions the party line seemed to be that MySQL was better because it could trash your data 10X faster than the competition.
Sure, I like having the option to speed things up by turning off transaction support. That is a wonderful feature to have. It is a horrible feature to require users to accept.
What I'm waiting for is transactional support at the filesystem level. Why can't I open a transaction, move a bunch of files around, edit configuration files to point to the new location, and commit the transaction? The best we can currently do is have a reasonable shot of having a single file in a somewhat consistent state after a crash. There really isn't any reason that file systems can't truly be transactional.
Couldn't agree more. My other big pet peeve is applications that are database-specific (which obviously is related to the fact that app designers don't have much choice in the matter). I don't want to install my own MSSQL server for the sake of the one application that can't run on my company's enterprise scale server farm running some other RDBMS. I don't want to pick a different application that doesn't meet my user's needs simply because it handles the other RDBMS.
Oh, and if we actually had some standards perhaps there would also be a chance that every time Oracle releaes a new software releaes it wouldn't break half the applications I manage...
That's only the problem from the point of view of the guy with the toxins.
And a problem for the person who can't buy a drug because everything it is made from is toxic. And a problem for people who buy stuff that can be made in other ways but which end up costing more, meaning that they have to choose between food, drugs, and their fancy new car. And so on...
Everything is connected. Anything that saves people money frees up resources for other things that are more important. If it costs grandma down the street $50 less per year to make her scrapbook albums then that is $50 more in gifts to her grandson who can buy a toy microscope, get interested in biology, and discover the cure for cancer. All things being equal it is better to develop the same stuff using fewer resources.
Any constraints we put on the use of technology raise costs. So, it is better to set constraints of "use this in a safe way" than "don't use this at all."
Again, I can't speak to the specific details of asbestos, but this principle applies in many areas of life, and as a society we tend to choose killing 1000 people statistically over killing 1 person more directly.
I didn't really want to spell out all the dirty details... I don't consider my TV critical - so I haven't invested in the resources to properly backup hundreds of GB of mpg files. I do consider a subset of my data critical, so I routinely back that data up to a separate system (which is located offsite much of the time, and easily carried with me if the house catches on fire).
So, when the data got hosed I just copied what I could to another computer (glitches and all), wiped everything down to the md drives, and then repartitioned. I then restored backups and also glitchy versions of stuff like TV where I can tollerate errors.
Like I said - I used to swear by LVM. In the right circumstances I would still use it. Those circumstances would include full backups of anything that I'd be annoyed at losing. I already miss some of the flexibility of LVM - I ended up making a few larger partitions on separate raid devices - I can always grow the filesystems if I enlarge my RAID, and now that my system is mostly full of drives I'm unlikely to move volumes around much.
This issue is obviously rare - no question about it. However, if you're a typical home user who isn't going to invest in 100% backup coverage it would give me pause. When you're talking about DVRs full backup coverage starts getting expensive, and I'm not sure anybody in my family really missed losing the odd TV episode - especially with re-runs all the time. Everything in/etc,/home, and stuff like mysql dumps were backed up.
I guess the other lesson here is ALWAYS BACK UP ANYTHING IMPORTANT. Fancy tools like software RAID, LVM snapshots, etc are all nice, but backups on completely separate media will protect you from a lot of logical errors.
In the meantime, I continue to wait for ZFS (which is not growable/shrinkable yet - that would be a necessary condition in addition to some maturity and preferably non-FUSE).
I can't say I know much about asbestos in particular but there are many substances in modern society that are unsafe if misused, but which have the potential to benefit life if they are used properly. Most industrial chemicals fall into this category - many are unsafe if mishandled, but if you start banning them left and right there won't be much of modern society left.
The question is one of risk vs reward. The problem is that in modern society we've become so risk adverse that we can't see beyond this. The other problem is that unscrupulous individuals cut corners and bend rules to make a buck. We really need to get beyond this - by reigning in both legal and corporate malfesance.
Nobody should have to live near a plant that releases toxins into the environment. However, the solution to this isn't to completely ban materials that have the potential to improve the general standard of living.
I would stick with ext3 - it is really the only option that meets your needs (which is why I'm using it as well). Note that I'd avoid using LVM - there is some kind of bug in some versions of LVM that causes massive data loss in some very rare circumstances. I recently lost a few hundred GB of data on a RAID due to this issue. (Google for "access beyond end of device lvm".) Ran fsck to clean up some errors after a crash while in RAID recovery mode and suddenly I had massive data loss on an entirely different lvm logical volume - it was obvious that the fsck somehow crossed the logical volume boundaries which should not be possible.
In the end I ended up restoring critical data from backups (which did not include mythtv recordings), and watched what remained of my recordings (complete with 10 second patches of video jumping between shows). I had to completely wipe out everything on the raid and start over. I no longer run lvm - I used to swear by it but it will be a while before I go back to it. My few non-lvm partitions (root, boot) had no issues at all even though they were subject to the same treatment.
You want the police to keep track of how long your unlicensed car has been on the road?
Yes.
If it is a serious problem (rare) then the police won't have trouble keeping track of it. In many areas police routinely tag abandoned cars and then tow then a few days later. It actually isn't that hard to do. If it is an eyesore a neighbor would report it.
My logic is that there should be a good reason for any particular law. If you want to say that the police should tow any parked car that lacks some piece of paper even for a second, you need to explain why that is important to a well-functioning society (as opposed to having a grace period).
Laws exist to benefit people - not to subjugate them.
A handful of people work because they want to. Most work because they have no choice. Do you think the greeters at Wal-Mart are there because they want to? Or because they have to eat?
This statement is true, has been true, and will probably always be true of all people at all times everywhere. It certainly is true for me and I'm nowhere near retirement.
Nobody "works" because they "want to". Sure - some people have the financial freedom to engage their time in activity of their own choice, but that isn't necessarily "work" per se.
The 30 year old guy hauling your garbage probably didn't aspire to this job since childhood.
Life isn't fun sometimes - we have to make do with reality.
I think that if you look back historically you won't really find "retirement" as a widespread practice in any society. This was mostly invented by the new deal, and under different terms than today (same age, but not relative to life expectancy).
I think the real crisis isn't so much that people don't get to sit back and relax for a decade or two at the end of their lives, but rather that due to life expectancy increasing faster than productive life expectancy we're creating a class of near human vegetables who need to be fed but can't do much else. If you could make humans healthy and strong up to the day the die then working your entire life wouldn't be such a problem.
Don't get me wrong - I'm sensitive to the plight of the elderly. The problem is that we're solving the wrong problem. The fact is that no society can function when a significant portion of its population expects to consume but not produce. That unfortunately is just reality - I didn't invent it and I'm sorry to be the bearer of bad news. I'm sure this isn't fair, but again, I didn't create the universe either.
If somebody can come up with a better system I'm all ears. Social security (as it currently exists) isn't it - and one way or another that will become apparent in another 10-20 years.
Look, if somebody wants to park a wreck on the side of the road and leave it there for two months I'm all for having it towed and the owner/dumper fined. These laws at some level make sense.
The problem is that laws are being written with zero-tolerence and razor-thin enforcement margins.
The purpose of laws is to prevent members of society from harming one another - they're rules that we can all live by (not die by). What harm did the guy cause by taking a week or two to have his car licensed/etc?
Live and let live is a much better way to get along. Sure, laws are needed and when somebody is being a jerk I'm all for hauling them to court. However, we don't need laws for the sake of themselves.
Still, it makes you wonder whether other astronomical events could have had an impact. Suppose some supernova nearby blasted the earth with neutrinos and caused 10% of a sample of isotope to decay in seconds - then return to normal rates of decay? Suppose the sun drifted into some cloud of dark matter a billion years ago and that messed things up?
We always assume the laws of physics are the same everywhere. This is probably true at a fundamental level, but it doesn't mean we understand all the laws of physics. A few hundred years ago you could have convinced a scientist that you could manipulate gravity (with hidden electromagnets). The reality isn't that gravity didn't work right, but rather that there was a previously-unknown force at work. There could be all kinds of fields at work in the universise that haven't been apparent to us simply because they're flat on the scale that we've experienced them. If a force only causes local effects but only within certain regions of the univerise and we're not in one of them then we'd never know it exists until the earth happens to pass into such a region and all kinds of stuff goes haywire.
I think a big problem is the net-present-value calculation when the expected rate of return of an investment starts hitting double-digit rates.
There is an expectation that money must always return at double-digit rates. MBAs are trying to turn EVERYTHING in to a growth company. They look at some start-up growing at 85% and think that a world-scale rubber manufacturer in an industry that has been mature for 30 years should be able to do the same.
When you expect these kinds of returns, then almost no investment at all looks profitable. Unless something pays for itself in the first year or two, it is considered a loss. Very few investments of any kind recoup all their principal within a few years.
The investments only get made when those pitching them are willing to lie about their true benefits. And when you're a CEO do you give your money to the manager that is promising an 8% return, or the guy who insists he can get 30% for it? I'm surprised these guys went to college at all and didn't just buy into a pyramid scheme...
The previous poster was right-on when mentioning a short-sighted view. Very few stock purchasers look beyond the next quarter for growth, and as a result very few managers to the same. When you can buy and sell almost for free, why would you invest for the long-term? If a company look like it will grow by leaps and bounds in a year, then you wait until six months from now before buying stock and put your money elsewhere for the time being. That mentality ends up translating into every department's quarterly budget.
Actually, the only reason that Bell did so much R&D was the way that utility regulation worked at the time.
Basically Bell would document its costs to deliver phone service, add a percentage, and present that to the regulatory bodies, who would approve new rates. The more Bell spent on delivering service, the more money it made.
If Bell burned $1000 in a fireplace and could argue that it was necessary to provide service, then the regulators would force consumers to cought up $1100. It was like printing money.
THAT is why they did so much blue-sky research. The fact is that if the goal were to do blue-sky research the money would have been better spent actually creating a lab for this purpose and cutting out the middle-man.
However, government is short-sited, so while research that helps fund kickbacks to monopolists is good, research that just cures cancer or otherwise benefits the public is a waste of tax dollars that could be better spent on stuff that garners more votes...
Yes, but what happens when they see 4 banner ads and then go buy the product at Walmart?
What happens when you see 14 ads, then a friend is asking you a question about a type of product and you mention that you saw an ad for such and such and the friend goes and buys it?
Impressions are impressions - they have an impact whether or not somebody ever buys anything at all. That is the problem with the way we measure ads.
This is just an example of the problem with online advertising.
The problem with online advertising is that it makes click-through rates very measurable. In the print world you never really know what the direct impact of an ad is (as in what the last ad somebody looked at was before they bought the car).
My thought has always been that the purpose of ads is to make a general impression on a viewer and keep a product in their mind. The Coke and Pepsi advertise, they aren't really competing against each other so much as the 47,000 generic brands of soda that somebody might otherwise buy if they didn't realize that there was something "special" about those two particular brands. Generic is generic because there isn't any advertising.
The problem on the web is that an ad isn't considered very effective unless soembody clicks on it. That is just silly. Nobody clicks on superbowl ads, and yet those are considered highly effective. Also - an online ad doesn't have the kind of durability that a print ad has. You can pick up a copy of the NYT and see that the ad did take up so many column-inches. With the NYT online the ad might not even be on the page at all when you happen to look at it - it just generates so many impressions per hour or whatever.
Online advertising is every bit as valuable as any other kind of advertising, but the fact that we can (mis)measure it so much more effectively makes it appear less valuable. I've heard a similar argument made as to why outsourcing is so popular. When you look at the department of 1000 people that you manage you can see what all the problems are and what your costs and limitations are. When you look at an outsourcer all you see is a cheaper cost and promised benefits. You've comparing reality against an ad - and then you act surprised when the glossy new car doesn't really do a better job of getting you to work than the 10-year-old car did.
I'm not completely certain, but I believe that since the Chinese government's official position is that Taiwan is just a renegade province they offer Chinese passports to anybody who asks for one. That gives Taiwanese easy access to mainland China. The more the Taiwanese take advantage of this the more China's claim is bolstered, so for them it is a political back-door.
It wouldn't surprise me if there are some limits on this, and I don't profess to know all the details. However, I do believe I read about the somewhere.
As far as democracy goes - I'm not sure that China needs an overnight revolution, but the argument that they're not quite ready yet could last 400 years. At some point you just need to start making changes.
Even if the government posted a 15 year plan to move to democracy it would probably be hailed as a good thing - especially if freedoms were granted early on, with clear education efforts in the plan to make democracy work. You could even use a progressive system in which voting rights are granted by education or income - with a new bracket being given the vote each year.
The problem is that right now the Chinese government makes itself an obvious target for criticism with the great firewall. That is like a relic from Stalinism...
Slashdot Mirror
I know a guy who works for a company that does a lot of lab testing of its products. Many years ago they were outsourcing quite a bit of testing to a number of companies and getting cheap prices. However, they had all kinds of quality issues - the various labs were not consistent with each other and the bottom-line mentality was making it hard to use the data that was generated.
So, they consolidated all their labwork to a single contract facility (the one that consistently had the best quality). Costs went up, but it was still cheaper than in-house, and the volume led to a partnership relationship with the testing lab, which was eager to compete on quality instead of price. Quality and throughput were better than ever, and cost was still very reasonable.
Fast forward a few years, and now the big rage is outsourcing. Now they're trying to go back to getting 5 vendors bidding against each other to squeeze every nickel of cost out of the system. How long do you think it will be before bad decisions are made on the basis of bad data?
When you get an excellent plasma TV for $2000 instead of $3000 you just saved $1000. When you get a broken plasma TV for $1500 instead of $2000 you didn't save $500 - you wasted $1500.
Maybe the ivory tower folks might not understand this, but perhaps the average American would rather be hung in a harness for 5 hours than to spend 30 hours on a train, or spending an extra $5000 for an airline ticket?
I agree that takeoff/landing slots should be limited to the quantity that can be safely handled. Probably the easiest way to handle this is via auction for 90% of the slots, with a fairly high fee for the last 10% so that non-advance-planned flights can still use the airport. Then consumers can choose whether they want to pay an extra $50 to be able to sit down on their flight. What's wrong with that?
The ultra-rich can still fly their private jets in complete comfort, although they would need to pay into the ATC maintenance like everybody else. They would pay a disproportionate share since the fees would be based on the cost to service an aircraft, and it is a heck of a lot cheaper per passenger to takeoff/land an A380 than a Gulfstream.
And maybe we can actually get to fixing the actual problem and stop putting band-aids over it. I'm not quite sure why humans are so heavily involved in the ATC system. You'd probably be far more efficient with an automated system that calls on humans when it needs help, and which relays instructions to planes digitally instead of making humans call out instructions and then read them back for verification. There is no reason that 95% of the time the computer in ATC can't just talk to the computer flying the plane, with humans available for failure modes and in case a little personal touch is needed.
Yup. Hence the reason I posted that in a discussion about the need to standardize SQL... :)
Somehow I doubt that will work with any frequency - you'd think newegg would wise up after the 3rd return with tool marks on it. If it did brick and mortar stores (which commonly have this policy) would be completely out of business.
If the item didn't have damage upon shipment, and it has a big bash in it when it is returned, then I'd claim for shipping insurance. If the insurer notices a trend in activity from a certain address then they're going to prosecute for insurance fraud.
There are lots of solutions to this problem. You don't need to stick it to your perfectly honest customers (the vast majority) just to get rid of a few abusers.
I agree - RMA for return for defective items should NOT have a restocking fee.
Sure - if the return is because "I don't like it" I'm fine with restocking fees. It helps a company like newegg keep costs down.
Also - returns should be parallel-ship - you ask for an RMA and the replacement item goes out immediately. If the return doesn't show up in x days charge the customer for a second item. None of that waiting around for mail to go back and forth.
And postage-paid returns are perfectly reasonable. It drives me nuts to have to pay to ship an item back if it is defective. I'm sorry - if a merchant gets defective junk they should tell their supplier to eat the costs - or threaten to drop the product. I'm Seagate would care if Newegg threatened to drop their product line. Sure, maybe it isn't cost-effective to test every item, but they can at least pay for the ones that leak through.
Most likely it is related to a motherboard that doesn't handle hard drive failures well - a dying hard drive was causing some flaky behavior. Even so, I can't see why that would cause LVM to cause data access beyond volume boundaries.
But I'll go ahead and echo your sentiments. I was running ext3+lvm+raid for a while without any issues at all. So are most people. However, I'll probably not do it again with data that I can't fully back up for a long time. Perhaps I'm just a lucky 1 out of 10,000. This is a DVR, so I'm not going to invest in what it takes to back up a few hundred GB of data offsite for some TV - nothing I lost wasn't expendible. It just was a real pain reinstalling everything.
Also - keep in mind that even enterprise grade hardware fails. The failure mode of the underlying software is still important.
I don't really see how this would be a distro issue. I was running versions of kernel+mdadm+lvm+ext3-tools that the upstream providers would have considered stable. There really isn't much else that should interact with the disks at that level.
I used fdisk to create partitions on the drives. Then I used mdadm to create RAID-5s on the partitions. Then I turned the /dev/md# raid devices into physical volumes and combined them into a volume group. Then I created logical volumes on top of that.
I can't see how the partitions couldn't have been valid. They start on one cylinder, and end on another. Not much more to it than that.
I didn't have any trouble with partitions that resided directly on md devices - such as boot/root. Only the logical volumes had problems. I also had a flaky hard drive at the time which was causing frequent crashes - once I figured out the cause the system was stable, but by then I had already hosed everything with an fsck. I figured that I couldn't go too wrong just running that on my video logical volume - after all that was all expendible. I didn't count on it corrupting stuff left and right on all my other logical volumes. With all the crashes the system did spend quite a bit of time in recovery mode - which might not have helped - a crash in recovery mode can't necessarily be cleaned up at the RAID level - so damage could leak into the LVM layer. Even so, the sector mapping used by LVM should only change when adding new logical volumes, and it should never be in a state that would cause one partition to overwrite another. I can't see how a crash should result in LVM allowing writes outside of logical volume boundaries.
Actually - the issue isn't that myth was single-threaded. The problem is that the filesystem itself totally blocks everything when a delete is in progress.
Myth didn't just move the delete into a separate thread. It also turned the delete into a piece-by-piece truncation of the file. This way the filesystem is never deleting more than a small amount of data at a time and other stuff can get in beteween the deletes.
Go ahead and try creating two separate processes - have one keep reading from a file (typical playback process), and then have an entirely different process delete a 20GB file while that is running. Watch the playback choke (assuming it isn't all cached in RAM).
The last time I checked it was also not possible to reshape a raidz. As in - I have 5 drives in a RAID, and now I want to have 6 drives in a raid. Then, I have 6 250GB drives in a RAID and I want to switch to 6 500GB drives in the RAID and get all the space without resorting to splitting the drives into 2x250GB partitions. I'm not talkinga bout adding another raidz to a pool containing a raidz.
Linux software RAID5 supports reshaping in this way (I think it even supports it online for the brave).
Don't get me wrong - I love the whole idea of ZFS and I think that copy-on-write will be the future of filesystems. However, the implementation isn't as mature as other stuff out there quite yet. I'm sure we'll get there eventually - there is no theoretical reason you couldn't reshape a raidz at least offline.
I'll vote for the "doesn't think it is important" option. That was the one thing that always annoyed me about MySQL. Rather than apologize for taking a while to support transactions the party line seemed to be that MySQL was better because it could trash your data 10X faster than the competition.
Sure, I like having the option to speed things up by turning off transaction support. That is a wonderful feature to have. It is a horrible feature to require users to accept.
What I'm waiting for is transactional support at the filesystem level. Why can't I open a transaction, move a bunch of files around, edit configuration files to point to the new location, and commit the transaction? The best we can currently do is have a reasonable shot of having a single file in a somewhat consistent state after a crash. There really isn't any reason that file systems can't truly be transactional.
Couldn't agree more. My other big pet peeve is applications that are database-specific (which obviously is related to the fact that app designers don't have much choice in the matter). I don't want to install my own MSSQL server for the sake of the one application that can't run on my company's enterprise scale server farm running some other RDBMS. I don't want to pick a different application that doesn't meet my user's needs simply because it handles the other RDBMS.
Oh, and if we actually had some standards perhaps there would also be a chance that every time Oracle releaes a new software releaes it wouldn't break half the applications I manage...
That's only the problem from the point of view of the guy with the toxins.
And a problem for the person who can't buy a drug because everything it is made from is toxic. And a problem for people who buy stuff that can be made in other ways but which end up costing more, meaning that they have to choose between food, drugs, and their fancy new car. And so on...
Everything is connected. Anything that saves people money frees up resources for other things that are more important. If it costs grandma down the street $50 less per year to make her scrapbook albums then that is $50 more in gifts to her grandson who can buy a toy microscope, get interested in biology, and discover the cure for cancer. All things being equal it is better to develop the same stuff using fewer resources.
Any constraints we put on the use of technology raise costs. So, it is better to set constraints of "use this in a safe way" than "don't use this at all."
Again, I can't speak to the specific details of asbestos, but this principle applies in many areas of life, and as a society we tend to choose killing 1000 people statistically over killing 1 person more directly.
Actually - I wasn't using snapshotting at all.
I didn't really want to spell out all the dirty details... I don't consider my TV critical - so I haven't invested in the resources to properly backup hundreds of GB of mpg files. I do consider a subset of my data critical, so I routinely back that data up to a separate system (which is located offsite much of the time, and easily carried with me if the house catches on fire).
So, when the data got hosed I just copied what I could to another computer (glitches and all), wiped everything down to the md drives, and then repartitioned. I then restored backups and also glitchy versions of stuff like TV where I can tollerate errors.
Like I said - I used to swear by LVM. In the right circumstances I would still use it. Those circumstances would include full backups of anything that I'd be annoyed at losing. I already miss some of the flexibility of LVM - I ended up making a few larger partitions on separate raid devices - I can always grow the filesystems if I enlarge my RAID, and now that my system is mostly full of drives I'm unlikely to move volumes around much.
This issue is obviously rare - no question about it. However, if you're a typical home user who isn't going to invest in 100% backup coverage it would give me pause. When you're talking about DVRs full backup coverage starts getting expensive, and I'm not sure anybody in my family really missed losing the odd TV episode - especially with re-runs all the time. Everything in /etc, /home, and stuff like mysql dumps were backed up.
I guess the other lesson here is ALWAYS BACK UP ANYTHING IMPORTANT. Fancy tools like software RAID, LVM snapshots, etc are all nice, but backups on completely separate media will protect you from a lot of logical errors.
In the meantime, I continue to wait for ZFS (which is not growable/shrinkable yet - that would be a necessary condition in addition to some maturity and preferably non-FUSE).
I saw that. But I'm running 2.6.25-gentoo-r7 - which is obviously newer than that. Could be a regression of some kind...
I can't say I know much about asbestos in particular but there are many substances in modern society that are unsafe if misused, but which have the potential to benefit life if they are used properly. Most industrial chemicals fall into this category - many are unsafe if mishandled, but if you start banning them left and right there won't be much of modern society left.
The question is one of risk vs reward. The problem is that in modern society we've become so risk adverse that we can't see beyond this. The other problem is that unscrupulous individuals cut corners and bend rules to make a buck. We really need to get beyond this - by reigning in both legal and corporate malfesance.
Nobody should have to live near a plant that releases toxins into the environment. However, the solution to this isn't to completely ban materials that have the potential to improve the general standard of living.
I would stick with ext3 - it is really the only option that meets your needs (which is why I'm using it as well). Note that I'd avoid using LVM - there is some kind of bug in some versions of LVM that causes massive data loss in some very rare circumstances. I recently lost a few hundred GB of data on a RAID due to this issue. (Google for "access beyond end of device lvm".) Ran fsck to clean up some errors after a crash while in RAID recovery mode and suddenly I had massive data loss on an entirely different lvm logical volume - it was obvious that the fsck somehow crossed the logical volume boundaries which should not be possible.
In the end I ended up restoring critical data from backups (which did not include mythtv recordings), and watched what remained of my recordings (complete with 10 second patches of video jumping between shows). I had to completely wipe out everything on the raid and start over. I no longer run lvm - I used to swear by it but it will be a while before I go back to it. My few non-lvm partitions (root, boot) had no issues at all even though they were subject to the same treatment.
You want the police to keep track of how long your unlicensed car has been on the road?
Yes.
If it is a serious problem (rare) then the police won't have trouble keeping track of it. In many areas police routinely tag abandoned cars and then tow then a few days later. It actually isn't that hard to do. If it is an eyesore a neighbor would report it.
My logic is that there should be a good reason for any particular law. If you want to say that the police should tow any parked car that lacks some piece of paper even for a second, you need to explain why that is important to a well-functioning society (as opposed to having a grace period).
Laws exist to benefit people - not to subjugate them.
A handful of people work because they want to. Most work because they have no choice. Do you think the greeters at Wal-Mart are there because they want to? Or because they have to eat?
This statement is true, has been true, and will probably always be true of all people at all times everywhere. It certainly is true for me and I'm nowhere near retirement.
Nobody "works" because they "want to". Sure - some people have the financial freedom to engage their time in activity of their own choice, but that isn't necessarily "work" per se.
The 30 year old guy hauling your garbage probably didn't aspire to this job since childhood.
Life isn't fun sometimes - we have to make do with reality.
I think that if you look back historically you won't really find "retirement" as a widespread practice in any society. This was mostly invented by the new deal, and under different terms than today (same age, but not relative to life expectancy).
I think the real crisis isn't so much that people don't get to sit back and relax for a decade or two at the end of their lives, but rather that due to life expectancy increasing faster than productive life expectancy we're creating a class of near human vegetables who need to be fed but can't do much else. If you could make humans healthy and strong up to the day the die then working your entire life wouldn't be such a problem.
Don't get me wrong - I'm sensitive to the plight of the elderly. The problem is that we're solving the wrong problem. The fact is that no society can function when a significant portion of its population expects to consume but not produce. That unfortunately is just reality - I didn't invent it and I'm sorry to be the bearer of bad news. I'm sure this isn't fair, but again, I didn't create the universe either.
If somebody can come up with a better system I'm all ears. Social security (as it currently exists) isn't it - and one way or another that will become apparent in another 10-20 years.
Uh, isn't this a case of laws being out-of-touch?
Look, if somebody wants to park a wreck on the side of the road and leave it there for two months I'm all for having it towed and the owner/dumper fined. These laws at some level make sense.
The problem is that laws are being written with zero-tolerence and razor-thin enforcement margins.
The purpose of laws is to prevent members of society from harming one another - they're rules that we can all live by (not die by). What harm did the guy cause by taking a week or two to have his car licensed/etc?
Live and let live is a much better way to get along. Sure, laws are needed and when somebody is being a jerk I'm all for hauling them to court. However, we don't need laws for the sake of themselves.
Still, it makes you wonder whether other astronomical events could have had an impact. Suppose some supernova nearby blasted the earth with neutrinos and caused 10% of a sample of isotope to decay in seconds - then return to normal rates of decay? Suppose the sun drifted into some cloud of dark matter a billion years ago and that messed things up?
We always assume the laws of physics are the same everywhere. This is probably true at a fundamental level, but it doesn't mean we understand all the laws of physics. A few hundred years ago you could have convinced a scientist that you could manipulate gravity (with hidden electromagnets). The reality isn't that gravity didn't work right, but rather that there was a previously-unknown force at work. There could be all kinds of fields at work in the universise that haven't been apparent to us simply because they're flat on the scale that we've experienced them. If a force only causes local effects but only within certain regions of the univerise and we're not in one of them then we'd never know it exists until the earth happens to pass into such a region and all kinds of stuff goes haywire.
I think a big problem is the net-present-value calculation when the expected rate of return of an investment starts hitting double-digit rates.
There is an expectation that money must always return at double-digit rates. MBAs are trying to turn EVERYTHING in to a growth company. They look at some start-up growing at 85% and think that a world-scale rubber manufacturer in an industry that has been mature for 30 years should be able to do the same.
When you expect these kinds of returns, then almost no investment at all looks profitable. Unless something pays for itself in the first year or two, it is considered a loss. Very few investments of any kind recoup all their principal within a few years.
The investments only get made when those pitching them are willing to lie about their true benefits. And when you're a CEO do you give your money to the manager that is promising an 8% return, or the guy who insists he can get 30% for it? I'm surprised these guys went to college at all and didn't just buy into a pyramid scheme...
The previous poster was right-on when mentioning a short-sighted view. Very few stock purchasers look beyond the next quarter for growth, and as a result very few managers to the same. When you can buy and sell almost for free, why would you invest for the long-term? If a company look like it will grow by leaps and bounds in a year, then you wait until six months from now before buying stock and put your money elsewhere for the time being. That mentality ends up translating into every department's quarterly budget.
Actually, the only reason that Bell did so much R&D was the way that utility regulation worked at the time.
Basically Bell would document its costs to deliver phone service, add a percentage, and present that to the regulatory bodies, who would approve new rates. The more Bell spent on delivering service, the more money it made.
If Bell burned $1000 in a fireplace and could argue that it was necessary to provide service, then the regulators would force consumers to cought up $1100. It was like printing money.
THAT is why they did so much blue-sky research. The fact is that if the goal were to do blue-sky research the money would have been better spent actually creating a lab for this purpose and cutting out the middle-man.
However, government is short-sited, so while research that helps fund kickbacks to monopolists is good, research that just cures cancer or otherwise benefits the public is a waste of tax dollars that could be better spent on stuff that garners more votes...
Yes, but what happens when they see 4 banner ads and then go buy the product at Walmart?
What happens when you see 14 ads, then a friend is asking you a question about a type of product and you mention that you saw an ad for such and such and the friend goes and buys it?
Impressions are impressions - they have an impact whether or not somebody ever buys anything at all. That is the problem with the way we measure ads.
This is just an example of the problem with online advertising.
The problem with online advertising is that it makes click-through rates very measurable. In the print world you never really know what the direct impact of an ad is (as in what the last ad somebody looked at was before they bought the car).
My thought has always been that the purpose of ads is to make a general impression on a viewer and keep a product in their mind. The Coke and Pepsi advertise, they aren't really competing against each other so much as the 47,000 generic brands of soda that somebody might otherwise buy if they didn't realize that there was something "special" about those two particular brands. Generic is generic because there isn't any advertising.
The problem on the web is that an ad isn't considered very effective unless soembody clicks on it. That is just silly. Nobody clicks on superbowl ads, and yet those are considered highly effective. Also - an online ad doesn't have the kind of durability that a print ad has. You can pick up a copy of the NYT and see that the ad did take up so many column-inches. With the NYT online the ad might not even be on the page at all when you happen to look at it - it just generates so many impressions per hour or whatever.
Online advertising is every bit as valuable as any other kind of advertising, but the fact that we can (mis)measure it so much more effectively makes it appear less valuable. I've heard a similar argument made as to why outsourcing is so popular. When you look at the department of 1000 people that you manage you can see what all the problems are and what your costs and limitations are. When you look at an outsourcer all you see is a cheaper cost and promised benefits. You've comparing reality against an ad - and then you act surprised when the glossy new car doesn't really do a better job of getting you to work than the 10-year-old car did.
I'm not completely certain, but I believe that since the Chinese government's official position is that Taiwan is just a renegade province they offer Chinese passports to anybody who asks for one. That gives Taiwanese easy access to mainland China. The more the Taiwanese take advantage of this the more China's claim is bolstered, so for them it is a political back-door.
It wouldn't surprise me if there are some limits on this, and I don't profess to know all the details. However, I do believe I read about the somewhere.
As far as democracy goes - I'm not sure that China needs an overnight revolution, but the argument that they're not quite ready yet could last 400 years. At some point you just need to start making changes.
Even if the government posted a 15 year plan to move to democracy it would probably be hailed as a good thing - especially if freedoms were granted early on, with clear education efforts in the plan to make democracy work. You could even use a progressive system in which voting rights are granted by education or income - with a new bracket being given the vote each year.
The problem is that right now the Chinese government makes itself an obvious target for criticism with the great firewall. That is like a relic from Stalinism...