if by mirroring you mean raid 1 then it protects against drive failure but not against accidental deletion or complete system failure (e.g. mains gets onto the DC rails and kills all the drives in your array at once) or complete building loss (building burns down).
If by mirroring you mean periodic mirroring to another drive which lives in the same machine or in another machine on the same. that can be a good soloution but the limited number of drives in a machine means you can only back up the most recent stuff if you insist on each copy being a complete bootable system mirror and you still have the problem of something killing all hardware in the system or even all hardware in the building.
If you mean mirroring to a removable or external drive which is safely stored and with at least one copy always in safe storage that can be a very good system but hard drives aren't really built to be carted arround all the time or to survive being in a fire safe that falls through several floors as the building burns down.
if you mean remote mirroring over a WAN that can be a great soloution if you have the bandwidth to handle it. Remote mirroring gets better but more expensive as you expand the geographical area and/or the protection of the mirror sites from any relavent natural disaters.
here are a number of bootloaders that can be used with both Linux and Windows including Lilo. Though I only have Windows on my PC for an OS I also have V Com's System Commander from which I could choose what OS I wanted to boot if I had more than one. sure there are but boot camp does more than that.
1: It allows windows to be installed without allowing the current bootloader to be clobbered (this one is likely to be hard to achive on PC hardware but at least a bootable CD can be provided that restores the linux bootloader. 2: It provides a CD full of drivers to make the windows install as painless as possible (not too hard, most OEMs already supply such a CD) 3: It handles the problem of adjusting partitions in a simple luser friendly manner. 4: It provides a series of step by step instructions for the entire process.
my guess is they will do what OEMs typically do for windows. Install a particular release (probablly one of the LTS releases) supply any drivers that are needed for thier hardware and don't come with that release and tell you that you are on your own if you upgrade to a new release. Same deal if you add new hardware your on your own.
one potential issue is the small ammount of support overlap. The lts releases are supposed to come out every 2 years and be supported for 3 years on the desktop (5 on servers). Therefore a machine sold just before a new lts release will only get just over a year of security and other maintinance updates.
Dell also has the option of working with ubuntu to ensure that new releases work out of the box on exiting dell machines.
there are still small phones availible, its just that they can't make them much smaller. Look at a modern small phone and you will find about half of it is the battery (battery technology is very mature) and much of the rest is stuff like keypad and screen (minimum size dominated by human factors). There just isn't much room left for making stuff smaller. Thus there are lots of phones in circulation that are as small as phones are likely to get and turn there is not that big a market for new ones.
thats true but his first comment still applies. its very unlikely your flash stick is formatted with a cluster size less than 4K anyway. Some of them actually don't use 512 byte sectors either (which can be a pain when booting off them).
If its a big flash stick it may have much bigger clusters, a 2 gig fat16 stick will have 32K clusters, a 4 gig fat16 stick will have 64K ones. I'm not sure if the larger flash sticks tend to come formatted fat16 or not but even if they are formatted as fat32 i highly doubt they will have clusters smaller than 4K.
Only thing i dunno, back on topic, is why on heaven shuttleworth is minding about the server market. Money! Shuttleworth is independently wealthy but I can think of a number of reasons for someone in his position not to want to support thier project indefinately (see list below) and the best place to make money in the linux market is "enterprise" contracts (basically providing support for systems that are very important to a company and hence have a large security budget associated).
1: customer confidance. What if shuttleworth was run over by a bus tomorrow? Would whoever inherits his fortunate continue propping up ubuntu in the way he does now? would he have just left the mony to ubuntu. What if he just loses interest in ubuntu. From a customers point of view a distro that only survives because of ongoing financial support from its founder doesn't seem a very safe bet.
2: his money is finite. I don't know how long he can keep supporting ubuntu at its present burn rate, but if ubuntu grows its costs will grow with it and if it doesn't have income to offset those costs it will become more and more of a financial burden for shuttleworth.
3: he may have other things he wants to do with his money once he gets ubuntu sorted.
however all but the smallest buisnesses tend to buy volume licenses and afaict all volume licenses come with serious downgrade rights (they've started giving some downgrade rights to OEM customers of the vista buisness editions too but its very limited in comparision to what volume license users get). If you buy a volume license for on of the buisness editions of vista i belive you can choose to install windows 95, windows 98, windows NT 4 workstation, windows 2000 proffessional or windows XP proffesional instead and remain within your license.
so buisnesses in general don't have to move to a new version of MS software until they want to move.
However home users don't get this option (unless they buy a buisness edition of vista OEM and have suitable XP media availible to install from) and i belive MS do this on purpose so that employees will pressure thier employer to upgrade rather than standardising on the same version they use at work.
Home users can buy retail of course to get an older version but since there are no explicit downgrade versions this is a very expensive option. Alternatively they can pirate stuff but MS has been making life harder for pirates recently (iirc wga actually has an option for snitching on pirates in exchange for a free legit copy).
People are certainly more willing to go back to XP, but pretty soon they won't have the choice. they can always buy the buisness version of vista and excercise the downgrade rights that come with it assuming that they have access to suitable media to install from.
I can't find any evidence of the realtek comment in the linux tree and since the driver its from is has a four clause BSD it seems unlikely it was ever in linux. It does seem to be in all 3 of the major BSD variants though.
In realteks defense mind you i think the worst PCI ethernet controller comment is a bit unfair, sure its a bit heavy on the CPU and annoying to drive but it seems to be a pretty stable chip and its damn cheap.
TBH i suspect you will find lots of comments about hardware issues in any driver source unless its been through a comment stripper. Hardware has bugs and design flaws that can't really be fixed in the hardware either due to the cost of a respin or due to the fact that units are already in the wild. Some of those bugs will be in the IC makers errata but getting them to add newly found issues to the errata can be very hard. I strongly suspect the only difference is that linux/freebsd make those comments fully public whereas with propietry operating systems you'd probablly need a source license to see them.
its not as though the governement can't turn up on a private ISPs doorstep with mass wiretapping orders anyway or indeed just spy on it themselves (assuming its not encrypted or they get a mole who can get the keys).
the problem with subscriptions is you tend to end up locked in.
What if you end up a bit poorer? if you were buying music you'd just stop buying more and coast along with what you've got, if you were renting you'd lose your music collection completely until you subscribed again.
what if your provider cranks up prices and you see a better deal elsewhere but you really can't face the thought of re-finding and re-downloading your entire collection or you simply don't have the bandwidth to re-download it in a reasonbale time?
what if you are looking get a new portable player. don't you want the full choice of the market rather than being tied to a smaller selection of player makers (granted this applies to online buying too right now but that looks set to change whereas renting looks like it will always be drm). What if you want to switch OS, possiblly to one which doesn't support drm?
Some things you can only get by renting them because you are effectively paying for services that go with them (communications links, public utility connections) or because you don't have the capital upfront (housing) but for the most part renting is something to be avoided imo.
People who value education oppose schooling, because schooling does very little education. Modern schools are indoctrination and daycare centers which keep kids out of the way for fifteen years while their parents are working... any actual education is an unwanted side-product. And what do you propose instead?
Schools may have problems but far far more people can read and write to a reasonable level and have at least some maths knowlage than in the days before compulsory schooling. If children didn't spend that time in school they'd spend in hanging arround doing little of consequence and most likely getting into trouble.
or they'd be working in a dead end job in a factory where they wouldn't really be mature enough to work safely.
no we have spam because there is little in the way of consequences for those whoose computers are used to send it. At worst they get put on a blocklist that means they have to move to another IP. At best they don't have any noticable consequences at all because the spam is sent out in a different way to thier regular mail and there is not enough traffic to noticablly impact thier bandwidth.
imagine someone started using your credit card to pay to send out thier physical junk mail, you'd and your credit card company would be trying to stop them pretty damn quick. OTOH if someone uses your computer to send spam you probablly won't notice until someone complains and possiblly not even then.
you don't explain that until later. Its a hack that some people will need to be made aware of in explaining why stuff doesn't work but its not fundamental to the internet.
I'd also question whether a machine without a public IP was actually on the internet in the first place, imo there is a difference between "has access to the internet" and "on the internet" though the distinction is a bit blurry.
The difficult part is to explain why you pay so much for a phonecall two reasons 1: Phone is higher priority (both in terms of priority in bandwidth and in terms of fixing stuff quickly when it goes wrong) 2: monopolistic practices
with voip you can call anyone on your voip network for nothing more than the cost of bandwidth from your ISP and you can call people on POTs in the western world pretty cheap too but things may be unreliable and use of it will be blocked in some places by monopolistic local governemnts.
it depends what you count as a phone line, afaict most internet traffic runs over the same types of trunk lines that were originally designed for phone.
that will stop cartridge cloners, it won't stop builders of inkjet PRINTERS.
patents can make building PRINTERS much harder (witness the fact that there are only a few makers atm) but as the GP said patents have a finite lifetime.
the cartridge chip thing only affects builders of third party carts for existing printers.
indeed the general purpose userspace API/ABI is pretty stable though newer glibc versions have started to depend on 2.6 (most notablly NPTL), I assumed that the broke the ABI comment was in reference to the modules APIs.
your best bet IMO if you don't want to give the customers any special kit is to host a largeish file on a powerfull server with a good server class PCI-X or PCI-E gigabit ethernet card. Then get them to download it from multiple machines at once. You can then measure the traffic going out of the server (either with software on the server or with your existing infrastructure equipment).
Afaict (i don't have one but i'm thinking of getting one) the intel macs are good quality predictable (you know what your getting/its easy to research linux support) hardware and they can run mac OS X which many people think is one of the best desktop operating systems arround (lets be honest, the linux desktop is a mess of many different widget sets and poor integration) and has very good (some would say better than windows, it certainly used to be that way though increasingly software companies have been jumping ship) software support for certain arty niches (including probablly the one you mention).
obviously hardware support will be needed for safe hotplugging but i don't see why it should be any less feasible to hotplug a CPU than to hotplug ram or expansion cards.
hotplugging is very important to those who run very high uptime systems, to meet thier stated uptimes some people basically can only afford to shut down every few years, in that time a fair bit of hardware can fail or need upgrading!
In general no, the ABI was not broken between previous releases (although the API has been broken numerous times). i'm pretty sure the policy has always been that they don't care about a consistant abi even accross minor releases, i'm sure linus even came out and said so.
if the bits of the abi your modules care about didn't change that was simply a side effect of them not tweaking anything too much in that area.
if an api is broken then the corresponding abi is almost certainly broken too.
personally i hate using an initrd.img and prefer to build ext2 & ext3 support right in the kernel making initrd unnecessary, if you compile file system support as a module you will need an initrd.img too so insetead of selecting an "M" select "*" you could try that... If it was just a matter of filesystem support i doubt the distros would have gone down the initrd root in the first place, hell doesn't ext2 have to be compiled in anyway for some types of initrd to work.
the real issue is hardware support, the bootloader uses bios calls but once the linux kernel proper is loaded its not easilly possible to do that (it could be done but i don't think anyone has ever considered it worth writing a driver that does access windows 95 compatibility mode style by repeatedly switching back to real mode). The result is that the linux kernel has to access the hard drive without using anything that is not already in ram. There are basically two ways to do this, either you build in the drivers for the hard drive (impractical for pre-built distro kernels due to the sheer number of possible drivers) or you use an initrd (which is placed in ram by the bootloader before linux is loaded).
Slashdot Mirror
if by mirroring you mean raid 1 then it protects against drive failure but not against accidental deletion or complete system failure (e.g. mains gets onto the DC rails and kills all the drives in your array at once) or complete building loss (building burns down).
If by mirroring you mean periodic mirroring to another drive which lives in the same machine or in another machine on the same. that can be a good soloution but the limited number of drives in a machine means you can only back up the most recent stuff if you insist on each copy being a complete bootable system mirror and you still have the problem of something killing all hardware in the system or even all hardware in the building.
If you mean mirroring to a removable or external drive which is safely stored and with at least one copy always in safe storage that can be a very good system but hard drives aren't really built to be carted arround all the time or to survive being in a fire safe that falls through several floors as the building burns down.
if you mean remote mirroring over a WAN that can be a great soloution if you have the bandwidth to handle it. Remote mirroring gets better but more expensive as you expand the geographical area and/or the protection of the mirror sites from any relavent natural disaters.
here are a number of bootloaders that can be used with both Linux and Windows including Lilo. Though I only have Windows on my PC for an OS I also have V Com's System Commander from which I could choose what OS I wanted to boot if I had more than one.
sure there are but boot camp does more than that.
1: It allows windows to be installed without allowing the current bootloader to be clobbered (this one is likely to be hard to achive on PC hardware but at least a bootable CD can be provided that restores the linux bootloader.
2: It provides a CD full of drivers to make the windows install as painless as possible (not too hard, most OEMs already supply such a CD)
3: It handles the problem of adjusting partitions in a simple luser friendly manner.
4: It provides a series of step by step instructions for the entire process.
my guess is they will do what OEMs typically do for windows. Install a particular release (probablly one of the LTS releases) supply any drivers that are needed for thier hardware and don't come with that release and tell you that you are on your own if you upgrade to a new release. Same deal if you add new hardware your on your own.
one potential issue is the small ammount of support overlap. The lts releases are supposed to come out every 2 years and be supported for 3 years on the desktop (5 on servers). Therefore a machine sold just before a new lts release will only get just over a year of security and other maintinance updates.
Dell also has the option of working with ubuntu to ensure that new releases work out of the box on exiting dell machines.
there are still small phones availible, its just that they can't make them much smaller. Look at a modern small phone and you will find about half of it is the battery (battery technology is very mature) and much of the rest is stuff like keypad and screen (minimum size dominated by human factors). There just isn't much room left for making stuff smaller. Thus there are lots of phones in circulation that are as small as phones are likely to get and turn there is not that big a market for new ones.
thats true but his first comment still applies. its very unlikely your flash stick is formatted with a cluster size less than 4K anyway. Some of them actually don't use 512 byte sectors either (which can be a pain when booting off them).
If its a big flash stick it may have much bigger clusters, a 2 gig fat16 stick will have 32K clusters, a 4 gig fat16 stick will have 64K ones. I'm not sure if the larger flash sticks tend to come formatted fat16 or not but even if they are formatted as fat32 i highly doubt they will have clusters smaller than 4K.
Only thing i dunno, back on topic, is why on heaven shuttleworth is minding about the server market.
Money! Shuttleworth is independently wealthy but I can think of a number of reasons for someone in his position not to want to support thier project indefinately (see list below) and the best place to make money in the linux market is "enterprise" contracts (basically providing support for systems that are very important to a company and hence have a large security budget associated).
1: customer confidance. What if shuttleworth was run over by a bus tomorrow? Would whoever inherits his fortunate continue propping up ubuntu in the way he does now? would he have just left the mony to ubuntu. What if he just loses interest in ubuntu. From a customers point of view a distro that only survives because of ongoing financial support from its founder doesn't seem a very safe bet.
2: his money is finite. I don't know how long he can keep supporting ubuntu at its present burn rate, but if ubuntu grows its costs will grow with it and if it doesn't have income to offset those costs it will become more and more of a financial burden for shuttleworth.
3: he may have other things he wants to do with his money once he gets ubuntu sorted.
however all but the smallest buisnesses tend to buy volume licenses and afaict all volume licenses come with serious downgrade rights (they've started giving some downgrade rights to OEM customers of the vista buisness editions too but its very limited in comparision to what volume license users get). If you buy a volume license for on of the buisness editions of vista i belive you can choose to install windows 95, windows 98, windows NT 4 workstation, windows 2000 proffessional or windows XP proffesional instead and remain within your license.
so buisnesses in general don't have to move to a new version of MS software until they want to move.
However home users don't get this option (unless they buy a buisness edition of vista OEM and have suitable XP media availible to install from) and i belive MS do this on purpose so that employees will pressure thier employer to upgrade rather than standardising on the same version they use at work.
Home users can buy retail of course to get an older version but since there are no explicit downgrade versions this is a very expensive option. Alternatively they can pirate stuff but MS has been making life harder for pirates recently (iirc wga actually has an option for snitching on pirates in exchange for a free legit copy).
People are certainly more willing to go back to XP, but pretty soon they won't have the choice.
they can always buy the buisness version of vista and excercise the downgrade rights that come with it assuming that they have access to suitable media to install from.
I can't find any evidence of the realtek comment in the linux tree and since the driver its from is has a four clause BSD it seems unlikely it was ever in linux. It does seem to be in all 3 of the major BSD variants though.
c /rtl81x9.c?rev=1.68&content-type=text/x-cvsweb-mar kup/ ic/rtl81x9.c
http://cvsweb.netbsd.org/bsdweb.cgi/src/sys/dev/i
http://crypto.riken.go.jp/pub/OpenBSD/src/sys/dev
http://fxr.watson.org/fxr/source/pci/if_rl.c
In realteks defense mind you i think the worst PCI ethernet controller comment is a bit unfair, sure its a bit heavy on the CPU and annoying to drive but it seems to be a pretty stable chip and its damn cheap.
TBH i suspect you will find lots of comments about hardware issues in any driver source unless its been through a comment stripper. Hardware has bugs and design flaws that can't really be fixed in the hardware either due to the cost of a respin or due to the fact that units are already in the wild. Some of those bugs will be in the IC makers errata but getting them to add newly found issues to the errata can be very hard. I strongly suspect the only difference is that linux/freebsd make those comments fully public whereas with propietry operating systems you'd probablly need a source license to see them.
its not as though the governement can't turn up on a private ISPs doorstep with mass wiretapping orders anyway or indeed just spy on it themselves (assuming its not encrypted or they get a mole who can get the keys).
the problem with subscriptions is you tend to end up locked in.
What if you end up a bit poorer? if you were buying music you'd just stop buying more and coast along with what you've got, if you were renting you'd lose your music collection completely until you subscribed again.
what if your provider cranks up prices and you see a better deal elsewhere but you really can't face the thought of re-finding and re-downloading your entire collection or you simply don't have the bandwidth to re-download it in a reasonbale time?
what if you are looking get a new portable player. don't you want the full choice of the market rather than being tied to a smaller selection of player makers (granted this applies to online buying too right now but that looks set to change whereas renting looks like it will always be drm). What if you want to switch OS, possiblly to one which doesn't support drm?
Some things you can only get by renting them because you are effectively paying for services that go with them (communications links, public utility connections) or because you don't have the capital upfront (housing) but for the most part renting is something to be avoided imo.
People who value education oppose schooling, because schooling does very little education. Modern schools are indoctrination and daycare centers which keep kids out of the way for fifteen years while their parents are working... any actual education is an unwanted side-product.
And what do you propose instead?
Schools may have problems but far far more people can read and write to a reasonable level and have at least some maths knowlage than in the days before compulsory schooling. If children didn't spend that time in school they'd spend in hanging arround doing little of consequence and most likely getting into trouble.
or they'd be working in a dead end job in a factory where they wouldn't really be mature enough to work safely.
no we have spam because there is little in the way of consequences for those whoose computers are used to send it. At worst they get put on a blocklist that means they have to move to another IP. At best they don't have any noticable consequences at all because the spam is sent out in a different way to thier regular mail and there is not enough traffic to noticablly impact thier bandwidth.
imagine someone started using your credit card to pay to send out thier physical junk mail, you'd and your credit card company would be trying to stop them pretty damn quick. OTOH if someone uses your computer to send spam you probablly won't notice until someone complains and possiblly not even then.
you don't explain that until later. Its a hack that some people will need to be made aware of in explaining why stuff doesn't work but its not fundamental to the internet.
I'd also question whether a machine without a public IP was actually on the internet in the first place, imo there is a difference between "has access to the internet" and "on the internet" though the distinction is a bit blurry.
The difficult part is to explain why you pay so much for a phonecall
two reasons
1: Phone is higher priority (both in terms of priority in bandwidth and in terms of fixing stuff quickly when it goes wrong)
2: monopolistic practices
with voip you can call anyone on your voip network for nothing more than the cost of bandwidth from your ISP and you can call people on POTs in the western world pretty cheap too but things may be unreliable and use of it will be blocked in some places by monopolistic local governemnts.
it depends what you count as a phone line, afaict most internet traffic runs over the same types of trunk lines that were originally designed for phone.
old black rotary phones
with mechanised devices turning them to dial and an accoustic coupler to pass the signals.
that will stop cartridge cloners, it won't stop builders of inkjet PRINTERS.
patents can make building PRINTERS much harder (witness the fact that there are only a few makers atm) but as the GP said patents have a finite lifetime.
the cartridge chip thing only affects builders of third party carts for existing printers.
indeed the general purpose userspace API/ABI is pretty stable though newer glibc versions have started to depend on 2.6 (most notablly NPTL), I assumed that the broke the ABI comment was in reference to the modules APIs.
and i bet most PCs aren't capable of it either.
your best bet IMO if you don't want to give the customers any special kit is to host a largeish file on a powerfull server with a good server class PCI-X or PCI-E gigabit ethernet card. Then get them to download it from multiple machines at once. You can then measure the traffic going out of the server (either with software on the server or with your existing infrastructure equipment).
what cleaning soloution do you dunk it in?
Afaict (i don't have one but i'm thinking of getting one) the intel macs are good quality predictable (you know what your getting/its easy to research linux support) hardware and they can run mac OS X which many people think is one of the best desktop operating systems arround (lets be honest, the linux desktop is a mess of many different widget sets and poor integration) and has very good (some would say better than windows, it certainly used to be that way though increasingly software companies have been jumping ship) software support for certain arty niches (including probablly the one you mention).
obviously hardware support will be needed for safe hotplugging but i don't see why it should be any less feasible to hotplug a CPU than to hotplug ram or expansion cards.
hotplugging is very important to those who run very high uptime systems, to meet thier stated uptimes some people basically can only afford to shut down every few years, in that time a fair bit of hardware can fail or need upgrading!
In general no, the ABI was not broken between previous releases (although the API has been broken numerous times).
i'm pretty sure the policy has always been that they don't care about a consistant abi even accross minor releases, i'm sure linus even came out and said so.
if the bits of the abi your modules care about didn't change that was simply a side effect of them not tweaking anything too much in that area.
if an api is broken then the corresponding abi is almost certainly broken too.
personally i hate using an initrd.img and prefer to build ext2 & ext3 support right in the kernel making initrd unnecessary, if you compile file system support as a module you will need an initrd.img too so insetead of selecting an "M" select "*" you could try that...
If it was just a matter of filesystem support i doubt the distros would have gone down the initrd root in the first place, hell doesn't ext2 have to be compiled in anyway for some types of initrd to work.
the real issue is hardware support, the bootloader uses bios calls but once the linux kernel proper is loaded its not easilly possible to do that (it could be done but i don't think anyone has ever considered it worth writing a driver that does access windows 95 compatibility mode style by repeatedly switching back to real mode). The result is that the linux kernel has to access the hard drive without using anything that is not already in ram. There are basically two ways to do this, either you build in the drivers for the hard drive (impractical for pre-built distro kernels due to the sheer number of possible drivers) or you use an initrd (which is placed in ram by the bootloader before linux is loaded).