Hmmm, the Slashdot equivalent of the burglar complaining the homeowner shot at them while they were breaking in.
No. This is the Slashdot equivalent of shooting the tires of random package delivery trucks you see on the street,
because you strongly suspect some of the packages contain counterfeit copies of a product that you manufacture.
Run time optimizations can be (are?) a very bad idea in a kernel, where very often the exact predictability of execution paths makes the difference between a working kernel and a misbehaving one.
Huh? No.
If a code optimization causes a problem with the kernel, there is either a bug in the high-level code, where the constraints of the language have been violated by introducing an invalid runtime assumption, or a bug in the optimization procedure.
If specific code output were required, it should have been entered explicitly in the lower level language.
Untrue. Gcc is handicapped by political decisions in it technical design. It intentionally does not allow "others" to plug into some "internals". "Internals" that would facilitate other tool builders, especially those creating a graphical integrated development environment.
Software you can't plug-in to and interoperate with?
Isn't that the opposite of free software?
It raises the question... given the much better featureset, why are the major Linux distributions still using GCC to compile the kernel and all the OS packages?
He's just using the name Enterprise to try to get attention Things the enterprise has that his proposed ship does not:
Force fields (for maintenance of Hull Integrity, Navigation, Security, Medical Quarantine); Medical Stasis Fields; Metaphasic Shielding Emitters
Warp drive
Interstellar Navigation
Transporters
Tricorders
Turbolifts
Replicators, Holodecks, Phasers, Shields, Anti-Grav Units, Tractor Beams, Long range sensor arrays, Short range sensor arrays that can immediately identify life forms, objects, various substances, other ships, space/time anomolies
Actually they are NOT required to be consistent at all.
As long as they're not violating civil rights of a protected class, they can hire AND fire whoever they damn please.
Inconsistency has a danger -- if the employer fails to hire someone who conincidentally also happens to be a member of a protected class, and a complaint is raised; there will be a presumption of guilt, if the membership in the class is obvious to the interviewer, once the inconsistency comes to light, the employer's defense will be weak.
If the employer consistently asks/tests and rejects all candidates equally who have or lack certain non-protected characteristics, the defense will be stronger, "It's just their policy", in interviewing for this particular position; and the rejection can be shown to be unrelated to the protected class membership.
Hadoop is not useful for traditional applications.
Exchange's archiving system is not a replacement for traditional storage model,
for the users of Exchange
VMware sells a software Virtual Storage Appliance to turn a cluster of cheap servers into a nice storage device that you can use at a smallish scale.
It sounds like you missed the $5,000 software license fee for the VSA and haven't actually read the system requirements; the servers you can create a VSA configuration out of are not cheap servers, at least not anything normal businesses would call cheap.
After you add all the fees and hardware requirements; there are many traditional SAN options that are actually cheaper, more reliable, and more proven than the VSA.
The VSA is only valid on 3-node clusters; either a two node cluster, or a three node cluster.
If you create a 2-node cluster, you cannot add a third node later, without rebuilding from scratch.
You can't use a bunch of cheap desktops.
A minimum of 4 1-gigabit NICs per server is required for the VSA.
A bare minimum of 6GB of RAM is required per server to run the VSA,
and 24GB of RAM is recommended, just for the VSA; more RAM is required
since a VSA is useless without VMs on the servers.
RAID10 is the only supported configuration for the hard drives attached to a VSA.
RAID controller RAID controller must support RAID10
Only specific server models are supported, and they're not cheap.
It boggles my mind that Google has been lauging all the way to the bank on this issure for 10 years now, and as much as people try to copy Google in other areas, no one else even considers collections of commodity servers as a storage platforms (despite many open source glue solutions, even Google's own).
The problem is the "open source glue" solutions aren't very refined, have issues -- aren't production quality.
Google's system is not open source, and they have obviously expended massive efforts to develop it.
If your storage requirements and datasets are truly as massive as Google's, and the distribution of massive distribution of storage has parallel computing benefits you can tap, then it makes perfect sense.
Otherwise it's cost prohibitive.
The storage might be cheaper per-unit at massive scale.
But the cost to develop the technology is also massive -- and developing such a product takes a lot of time as well. It's not apparent at all that it's worth the cost for organizations that aren't in the storage business.
It does not make much sense to say that voltage "passes through" a resistor; it would be current.
It makes sense to say that a high voltage in the form of a static charge passes through a resistor, and discharges through a piece of electronics, destroying sensitive components such as semiconductor-based logic gates. When a resistor is subjected to a voltage on one side, a voltage difference builds at the input terminal of a resistor until the resistance is overcome by the difference in the potential at the output terminal, and then this difference must be equalized. There is a "drop" in voltage between the two terminals of a resistor, but this drop is not infinite, given a sufficient voltage on one side of the resistor -- the voltage will be discharged to the other side of that resistor,
there might be current in the form of a circuit, there might not be.
As far as lightning is concerned, a thick piece of plastic can be a resistor,
as can be plain air, and other materials that are normally conductors --
the voltage is certainly high enough, considering the lightning strike travelled
miles through the air to hit something near the soil... what's a few more feet?
Current divides based on the ratio of admittance, not resistance. Which, yes, is the inverse of ohms.
Electrical current isn't the only thing that matters. A static charge can still do a great deal of damage.
The characteristics of lightning discharge are variable, no matter how thick you make that copper, there is a possible lightning strike intensity that will defeat it, and you can't affordably make that copper thick enough.
Especially if the direct strike is inducing voltage in your inside wiring a significant distance away from your main panel.
(1) Manage your Facebook password with a password management application, so you can legitimately
tell them you "don't know" any Facebook passwords, they were randomly generated and are stored securely in a password manager at home and tied to that web browser.
Even if you have the correct FB password, you cannot login to FB on a new unknown device without receiving the SMS message, and entering the security code that Facebook sends you via SMS.
That's why it is illegal to ask a large number of questions that are not directly relevant to the job
It's not illegal. It's just inadvisable. If one of your questions exposes membership in a protected class,
there is a risk that there may be liability / possibility of a discrimination suit.
However ill advised, if you want to refuse employment to a janitor who never played
Chess or Checkers, can't remember the rules to the game Go, or can't beat the interviewer in a poker game,
the employer can do that, as long as they are consistent and require the same of every candidate.
It just happens that employers are in the business to select employees, and any irrelevent question
is a waste of time.
I'm not sure how I see this falling into the interstate commerce clause? I mean, a person works in his state....money paid to him in a state in which he is responsible for state taxes, etc.
Internet social networking sites have a multi-state presence; the federal government has
long claimed to have the jurisdiction over regulation of telecommunications services, see FCC.
In Facebook's case, it's a terms of use violation to share your password.
They are essentially passing legislation that forbids employers from interfering with
citizens' private relationship with certain other companies.
The legislation is broken though, because it's specific to social networking.
This should apply to all sites.
Including online banking sites, and sites where you pay your utility bills.
This is a form of consumer protection and privacy protection for interstate commerce.
Your private dealings are not your employer's business.
Your employer has no business seeing who your friends are, who your banks are,
what your account balances are, which cable package you subscribe, to,
what book you ordered from Barnes and Noble or Amazon,
what your viewing history is on Youtube and Netflix, etc.
And if some of employers are trying to pry anyways and demand passwords to
personal accounts their company has no right to, then it certainly is the feds' job to reign in the abuse.
You ALSO have to run bare copper back to the service entrance for water, and bond to that as well. In case one or the other fails, you still have a reliable path to ground.
That's not why you have to do it; "grounding" is relative and not a magic thing that guarantees no current will flow, electrical current can flow between "ground" connections. Ground potential varies from place to place, 10 feet away, ground can be at a different potential. Geology, Electromagnetic interference, solar activity, lightning, electrical faults elsewhere, and other factors can further exacerbate the difference.
Bonding is required for the same reason that Neutral and Ground must be connected together at one place (the main service panel). If you do not have Plumbing Ground and Electrical ground bonded, you have different parts of your system connected to ground at different places ---- this means, the ground on your service panel can now be at a different electrical potential than your plumbing.
What this means, is that if something conductive touches both your plumbing, and something connected to the main panel ground (or neutral), current will flow through that conductive thing, to equalize the potential of the different grounds.
If that conductive thing is a human, this could very well mean that someone dies, because they touched the tap electrically connected to the plumbing, and a kitchen appliance with a metal chassis connected to neutral.
Therefore, the requirement is that you already have these bonded together with a low resistance path..
The bonding ensures that both systems are always at the same potential, so current does not flow between
Neutral or Main panel ground and your plumbing.
I should add that if you have iron pipes, you can get much of the same protection by grounding to the water pipes at the closest point to where they run underground.
The important thing is there must be a single ground, and it must be a low-resistance ground (copper wire sized appropriately and attached with the proper UL listed clamp for the application).
If you do have both conductive pipes and a grounding spike, you have two grounding electrodes.
All grounding electrodes that exist must be electrically bonded with your system ground at the main panel.
E.g. you can't have conductive pipes, but fail to have them bonded or connected to anything -- that
creates an electrocution hazard, if someone touches a metal object connected to the plumbing and
another metal object grounded to the main panel at the same time.
There must be only one ground in a system, and that single ground must be bonded to all grounding electrodes
which connect to anything in the building.
And that bonding should occur at the main service panel, which must be the place where Ground and Neutral/Common are also bonded.
This will still not protect against induced voltages. Electromagnetic waves can go right through walls, you see.
Replace your standard walls with an uninterrupted tight conductive wire mesh, to provide shielding, lining every wall from the ground and floor all the way to the top of the roof, and enclose it completely at the top, at the bottom, and on every side.
You don't install surge protectors to defend against a direct lightning strike --- there is no real defense against a direct lightning strike - you can make it less likely with a well-engineered lightning rod protection system. You install surge protectors to defend against currents induced by nearby lightning in your wiring. This is protection against damage lightning can cause without actually striking your wiring, or your building. If lightning strikes a mile away from you, and hits the ground, or a tree: this lightning can still induce currents in unshielded underground and overhead power and data cables. If there is no surge protection, the induced currents may destroy sensitive electronics such as computer power supplies.
In a direct hit situation, lightning hitting a surged protected circuit can easily arc through any surge protector; human safety is paramount in the design of surge protection and electrical systems, anyways, so there are always compromises anyways, that is, surge protectors still share a common ground with everything else, and a direct hit clamped to ground can effect everything else tied to that ground ---
Remember, with resistors in parallel, the amount of current is proportional to the resistance - the amount of voltage passing through the higher resistance path is not zero. Even if 99.9% of the lightning strike is clamped to ground, the 0.1% can still be 10000 volts.
No commercially available surge protector apparatus able to be fitted to a home electrical system and other utility lines entering a building with a price that is remotely affordable to the average homeowner is capable of providing remotely robust protection against a direct strike.
Just use surge protectors where needed that have an equipment replacement guarantee - and make sure you're protecting phone lines, TV cables, etc too, not just power.
The problem is surge strips are inconvenient, and you might temporarily plug something in that really should have one, but because surge strips are inconvenient you don't have one on every single outlet..
Surge strips also don't work for hard wired devices -- such as Insteon wall switches, InLineLinc, X10, and other devices that get hardwired, for automatic light control - which are sensitive to overvoltage, and a surge can knock them all out.
but these wall-mounted racks scream 'switch-only' racks to me. What are your experiences? Is it possible to put servers in racks like these, or should I find a 'both-side-accessible' rack instead?"
Indeed. For a wall-mounted rack, stick with half-depth components. Such as switches, routers, and patch panels.
And stay away from components that require full rack depth.
Possibly very small shallow-depth 1U servers, but make sure to get enough U to have spacing between components.
Since you're building the place, I would kind of suggest making the room large enough for full access. with minimum 3ft clearance front and back, and ensure adequate exhaust of hot air out the back and import of cool air in the front.
Consider using a metal cabinet.
Slashdot Mirror
Hmmm, the Slashdot equivalent of the burglar complaining the homeowner shot at them while they were breaking in.
No. This is the Slashdot equivalent of shooting the tires of random package delivery trucks you see on the street, because you strongly suspect some of the packages contain counterfeit copies of a product that you manufacture.
Run time optimizations can be (are?) a very bad idea in a kernel, where very often the exact predictability of execution paths makes the difference between a working kernel and a misbehaving one.
Huh? No.
If a code optimization causes a problem with the kernel, there is either a bug in the high-level code, where the constraints of the language have been violated by introducing an invalid runtime assumption, or a bug in the optimization procedure.
If specific code output were required, it should have been entered explicitly in the lower level language.
Untrue. Gcc is handicapped by political decisions in it technical design. It intentionally does not allow "others" to plug into some "internals". "Internals" that would facilitate other tool builders, especially those creating a graphical integrated development environment.
Software you can't plug-in to and interoperate with? Isn't that the opposite of free software?
It raises the question... given the much better featureset, why are the major Linux distributions still using GCC to compile the kernel and all the OS packages?
He's just using the name Enterprise to try to get attention Things the enterprise has that his proposed ship does not:
What's "insecure" about an Excel spreadsheet?
If you're already running windows, edit the file > Properties, click advanced "Encrypt" the file on the file server using Windows EFS.
Add the list of authorized users' certificates so only authorized users can decrypt the file.
Make sure to setup an EFS recovery certificate, export that, and back it up somewhere.
Actually they are NOT required to be consistent at all.
As long as they're not violating civil rights of a protected class, they can hire AND fire whoever they damn please.
Inconsistency has a danger -- if the employer fails to hire someone who conincidentally also happens to be a member of a protected class, and a complaint is raised; there will be a presumption of guilt, if the membership in the class is obvious to the interviewer, once the inconsistency comes to light, the employer's defense will be weak.
If the employer consistently asks/tests and rejects all candidates equally who have or lack certain non-protected characteristics, the defense will be stronger, "It's just their policy", in interviewing for this particular position; and the rejection can be shown to be unrelated to the protected class membership.
Hadoop is not useful for traditional applications.
Exchange's archiving system is not a replacement for traditional storage model, for the users of Exchange
VMware sells a software Virtual Storage Appliance to turn a cluster of cheap servers into a nice storage device that you can use at a smallish scale.
It sounds like you missed the $5,000 software license fee for the VSA and haven't actually read the system requirements; the servers you can create a VSA configuration out of are not cheap servers, at least not anything normal businesses would call cheap.
After you add all the fees and hardware requirements; there are many traditional SAN options that are actually cheaper, more reliable, and more proven than the VSA.
The VSA is only valid on 3-node clusters; either a two node cluster, or a three node cluster. If you create a 2-node cluster, you cannot add a third node later, without rebuilding from scratch. You can't use a bunch of cheap desktops.
A minimum of 4 1-gigabit NICs per server is required for the VSA.
A bare minimum of 6GB of RAM is required per server to run the VSA, and 24GB of RAM is recommended, just for the VSA; more RAM is required since a VSA is useless without VMs on the servers.
RAID10 is the only supported configuration for the hard drives attached to a VSA. RAID controller RAID controller must support RAID10
Only specific server models are supported, and they're not cheap.
It boggles my mind that Google has been lauging all the way to the bank on this issure for 10 years now, and as much as people try to copy Google in other areas, no one else even considers collections of commodity servers as a storage platforms (despite many open source glue solutions, even Google's own).
The problem is the "open source glue" solutions aren't very refined, have issues -- aren't production quality.
Google's system is not open source, and they have obviously expended massive efforts to develop it.
If your storage requirements and datasets are truly as massive as Google's, and the distribution of massive distribution of storage has parallel computing benefits you can tap, then it makes perfect sense.
Otherwise it's cost prohibitive. The storage might be cheaper per-unit at massive scale.
But the cost to develop the technology is also massive -- and developing such a product takes a lot of time as well. It's not apparent at all that it's worth the cost for organizations that aren't in the storage business.
You must be completely unfamiliar with government work. Email holds the same weight as blood when it comes to a little CYA.
This belongs in your organization's e-mail archiving system, not your INBOX.
It does not make much sense to say that voltage "passes through" a resistor; it would be current.
It makes sense to say that a high voltage in the form of a static charge passes through a resistor, and discharges through a piece of electronics, destroying sensitive components such as semiconductor-based logic gates. When a resistor is subjected to a voltage on one side, a voltage difference builds at the input terminal of a resistor until the resistance is overcome by the difference in the potential at the output terminal, and then this difference must be equalized. There is a "drop" in voltage between the two terminals of a resistor, but this drop is not infinite, given a sufficient voltage on one side of the resistor -- the voltage will be discharged to the other side of that resistor, there might be current in the form of a circuit, there might not be.
As far as lightning is concerned, a thick piece of plastic can be a resistor, as can be plain air, and other materials that are normally conductors -- the voltage is certainly high enough, considering the lightning strike travelled miles through the air to hit something near the soil... what's a few more feet?
Current divides based on the ratio of admittance, not resistance. Which, yes, is the inverse of ohms.
Electrical current isn't the only thing that matters. A static charge can still do a great deal of damage.
The characteristics of lightning discharge are variable, no matter how thick you make that copper, there is a possible lightning strike intensity that will defeat it, and you can't affordably make that copper thick enough.
Especially if the direct strike is inducing voltage in your inside wiring a significant distance away from your main panel.
Your employer usually wants to know what one of your bank accounts is so that they can pay you.
Last I checked you don't need to know someone's banking details to cut them a check.
And the employer doesn't discover their details when you deposit the paycheck from the employer.
It also solves the problem of one of your friends spoiling your movie night party by texting on their cell phone.
(1) Manage your Facebook password with a password management application, so you can legitimately tell them you "don't know" any Facebook passwords, they were randomly generated and are stored securely in a password manager at home and tied to that web browser.
(2) Enable Facebook 2-factor authentication with a second cell phone. Don't bring that phone to any job interview.
Even if you have the correct FB password, you cannot login to FB on a new unknown device without receiving the SMS message, and entering the security code that Facebook sends you via SMS.
That's why it is illegal to ask a large number of questions that are not directly relevant to the job
It's not illegal. It's just inadvisable. If one of your questions exposes membership in a protected class, there is a risk that there may be liability / possibility of a discrimination suit.
However ill advised, if you want to refuse employment to a janitor who never played Chess or Checkers, can't remember the rules to the game Go, or can't beat the interviewer in a poker game, the employer can do that, as long as they are consistent and require the same of every candidate.
It just happens that employers are in the business to select employees, and any irrelevent question is a waste of time.
I'm not sure how I see this falling into the interstate commerce clause? I mean, a person works in his state....money paid to him in a state in which he is responsible for state taxes, etc.
Internet social networking sites have a multi-state presence; the federal government has long claimed to have the jurisdiction over regulation of telecommunications services, see FCC. In Facebook's case, it's a terms of use violation to share your password.
They are essentially passing legislation that forbids employers from interfering with citizens' private relationship with certain other companies.
The legislation is broken though, because it's specific to social networking. This should apply to all sites.
Including online banking sites, and sites where you pay your utility bills. This is a form of consumer protection and privacy protection for interstate commerce.
Your private dealings are not your employer's business.
Your employer has no business seeing who your friends are, who your banks are, what your account balances are, which cable package you subscribe, to, what book you ordered from Barnes and Noble or Amazon, what your viewing history is on Youtube and Netflix, etc.
And if some of employers are trying to pry anyways and demand passwords to personal accounts their company has no right to, then it certainly is the feds' job to reign in the abuse.
You ALSO have to run bare copper back to the service entrance for water, and bond to that as well. In case one or the other fails, you still have a reliable path to ground.
That's not why you have to do it; "grounding" is relative and not a magic thing that guarantees no current will flow, electrical current can flow between "ground" connections. Ground potential varies from place to place, 10 feet away, ground can be at a different potential. Geology, Electromagnetic interference, solar activity, lightning, electrical faults elsewhere, and other factors can further exacerbate the difference.
Bonding is required for the same reason that Neutral and Ground must be connected together at one place (the main service panel). If you do not have Plumbing Ground and Electrical ground bonded, you have different parts of your system connected to ground at different places ---- this means, the ground on your service panel can now be at a different electrical potential than your plumbing.
What this means, is that if something conductive touches both your plumbing, and something connected to the main panel ground (or neutral), current will flow through that conductive thing, to equalize the potential of the different grounds.
If that conductive thing is a human, this could very well mean that someone dies, because they touched the tap electrically connected to the plumbing, and a kitchen appliance with a metal chassis connected to neutral.
Therefore, the requirement is that you already have these bonded together with a low resistance path.. The bonding ensures that both systems are always at the same potential, so current does not flow between Neutral or Main panel ground and your plumbing.
I should add that if you have iron pipes, you can get much of the same protection by grounding to the water pipes at the closest point to where they run underground.
The important thing is there must be a single ground, and it must be a low-resistance ground (copper wire sized appropriately and attached with the proper UL listed clamp for the application).
If you do have both conductive pipes and a grounding spike, you have two grounding electrodes.
All grounding electrodes that exist must be electrically bonded with your system ground at the main panel.
E.g. you can't have conductive pipes, but fail to have them bonded or connected to anything -- that creates an electrocution hazard, if someone touches a metal object connected to the plumbing and another metal object grounded to the main panel at the same time.
There must be only one ground in a system, and that single ground must be bonded to all grounding electrodes which connect to anything in the building.
And that bonding should occur at the main service panel, which must be the place where Ground and Neutral/Common are also bonded.
This will still not protect against induced voltages. Electromagnetic waves can go right through walls, you see.
Replace your standard walls with an uninterrupted tight conductive wire mesh, to provide shielding, lining every wall from the ground and floor all the way to the top of the roof, and enclose it completely at the top, at the bottom, and on every side.
You don't install surge protectors to defend against a direct lightning strike --- there is no real defense against a direct lightning strike - you can make it less likely with a well-engineered lightning rod protection system.
You install surge protectors to defend against currents induced by nearby lightning in your wiring. This is protection against damage lightning can cause without actually striking your wiring, or your building. If lightning strikes a mile away from you, and hits the ground, or a tree: this lightning can still induce currents in unshielded underground and overhead power and data cables. If there is no surge protection, the induced currents may destroy sensitive electronics such as computer power supplies.
In a direct hit situation, lightning hitting a surged protected circuit can easily arc through any surge protector; human safety is paramount in the design of surge protection and electrical systems, anyways, so there are always compromises anyways, that is, surge protectors still share a common ground with everything else, and a direct hit clamped to ground can effect everything else tied to that ground --- Remember, with resistors in parallel, the amount of current is proportional to the resistance - the amount of voltage passing through the higher resistance path is not zero. Even if 99.9% of the lightning strike is clamped to ground, the 0.1% can still be 10000 volts.
No commercially available surge protector apparatus able to be fitted to a home electrical system and other utility lines entering a building with a price that is remotely affordable to the average homeowner is capable of providing remotely robust protection against a direct strike.
Just use surge protectors where needed that have an equipment replacement guarantee - and make sure you're protecting phone lines, TV cables, etc too, not just power.
The problem is surge strips are inconvenient, and you might temporarily plug something in that really should have one, but because surge strips are inconvenient you don't have one on every single outlet..
Surge strips also don't work for hard wired devices -- such as Insteon wall switches, InLineLinc, X10, and other devices that get hardwired, for automatic light control - which are sensitive to overvoltage, and a surge can knock them all out.
Yes. Is is "accepted practice?" Yes. Just don't ask for a birth certificate!
Who's accepting the practice?
I'm not accepting the practice. It's not legal, it's totally unethical, and it's not OK.
The fraud should be referred to local authorities so the person doing that can be arrested, and he should be fired on the spot.
but these wall-mounted racks scream 'switch-only' racks to me. What are your experiences? Is it possible to put servers in racks like these, or should I find a 'both-side-accessible' rack instead?"
Indeed. For a wall-mounted rack, stick with half-depth components. Such as switches, routers, and patch panels. And stay away from components that require full rack depth.
Possibly very small shallow-depth 1U servers, but make sure to get enough U to have spacing between components.
Since you're building the place, I would kind of suggest making the room large enough for full access. with minimum 3ft clearance front and back, and ensure adequate exhaust of hot air out the back and import of cool air in the front. Consider using a metal cabinet.
Did you consider that Netflix is not the problem, it's the asinite ludicrously expensive bandwidth metering?