It's simple enough for me. I just looked at my system this weekend and realized I'm not even running LVM, and that it's a holy hell of stuff that just kind of fell into place because I haven't been picking it apart with tweezers. I've been running as a regular old user, using the software center and GUI tools for a long time; I even used the Software Center to install Steam by clicking the "Software" icon and typing "Steam".
At the same time, I'm designing a new swap daemon to manage swap on Linux. That whole swap partition thing is dumb now, because the installer on various distributions (notably Debian-ites) frequently allocates hundreds of gigabytes of swap when you have 64 or 96 gigabytes of RAM. The right solution would be an automatic swap manager using various strategies; my primary strategy is a zram swap strategy (allocate more zram devices, coalesce them, and otherwise increase zram as memory pressure increases until a fixed limit--say half of RAM is holding compressed data), and other workloads may need alternates such as zswap (uses swap files, and caches them in compressed memory--less than 100% of swap is in RAM), swap files, or LVM swap partitions created out of free volume group space on-the-fly.
A swap daemon like that would free the user from worrying about what the installer automatically does (e.g. eat 110GB of your 128GB SSD for a single swap partition), how much swap space they need up-front, and so forth. The extended configuration allows reconfiguration for specialized workloads (notably, scientific computing). In normal function, the computer would just behave better by default, OOMing less, struggling less when memory usage is high, and not consuming disk space for memory swap space; the user wouldn't even know about the new swap daemon unless someone told him.
That's what all this complex esoterism is: a bunch of shit that goes under the hood. Nerds care about it and we talk about it; end users only know that their graphics card actually works in this release.
It's out of the software's control, and still is completely relevant. The strategy of relying on shared pages makes the difference between whether a multiple-process model is *necessarily* huge bloat or easily-optimized to not bloat. For the application developer, it means designing the application to make some runtime state more stable (i.e. it's different per-run, but less-likely to change during a run) or to group strongly-stable state together (e.g. mmap() anonymous memory for blocks of things which repeat down to the child processes but don't change after they're allocated in the parent) reduces total physical memory usage, even though this may increase measured memory usage.
The software can't make it happen or detect it happening; it can account for it being likely.
Everyone does not struggle. There are reasons you have difficulty--not that you're too stupid for college, or that the next guy has a better brain; it's that you're using the wrong methods, and you're entering an unfamiliar environment.
The brain, first and foremost, is an energy-hungry organ. To minimize energy usage, it restructures to readily follow the most common set of actions. Overriding this--self-activation or response-inhibition--requires first formulating a plan in the prefrontal cortex, then activating the dorsolateral prefrontal cortex to directly control the midbrain. In short: your brain out of high school is a high school brain, and making it take actions different in *any* way from high school requires using one part of your brain to force the rest of it into compliance. Those parts of your brain all consume energy and, just as doing an endless stream of push-ups will eventually leave your muscles physically-incapable of lifting any more, your brain will run down the ATP supply in involved cells and become fatigued and, eventually, physically-incapable of taking these actions.
Many people work around this by setting schedules and reframing.
Setting schedules creates a low-energy activation system: your brain, over an average six weeks of effort, restructures such that keeping to that scheduled set of actions is a lower-energy behavior. Over years and with the experience of repeatedly changing schedules as necessary, your brain even develops the facility to set new schedules and adhere strongly to them in a few weeks as *the* lowest-energy behavior. Lowering the energy expenditure makes it easier to cope with something like a move from high school to college.
Reframing reduces energy expenditure by connecting one impulse to another. If you can evaluate an activity in terms of how it supports something you are actively interested in, that activity becomes a low-energy pursuit. Think of it like changing your work into a distraction: studying engineering is hard and boring, but you *like* tinkering with engines and building go karts and such; studying engineering will let you build *better* go karts, and that connection draws your brain to react on impulse by studying engineering. When that study converts quickly and recognizably into a greater depth of enjoyment of your hobby, you get positive reinforcement, causing your brain to structure itself around seeking those results by executing those actions.
Besides all of this, the geniuses among us are only using competent mental techniques. These range from scheduling and reframing as above to structured study techniques (SQW4R, OK4R), casual or systemic mnemonics (simple visualization up through mind palaces and other systems), and even structured note-taking systems (Cornell notes are the best generic; there are better systems for certain specific materials). Developing and exercising the correct skills *vastly* reduces the difficulties faced in college because time spent studying is optimized: less effort and less time produces more complete understanding and better grades.
Telling people to grin and bear it because college is hard is idiotic. If the effort and cost of college are high enough, it's *not* worth the investment--I dropped out and it was the best decision of my life; this may have been wholly untrue if I had employed more-advanced study techniques (and, in fact, I've gone back and structured my own computer science program to self-teach on the weekends because it's a six-month task if I can find 1 hour per day). Identify *why* they're struggling and fix it.
The built-in browser tools know which virtual pages are mapped to the same physical pages, even though knowing that information would require getting kernel-level access to read the page table entries for the process? Does Chrome install a kernel module?
The amount of RAM used by a set of processes which takes the exact same actions (down to the areas of memory allocated and the exact memory addresses used for every byte of data) will be different between Linux and Windows.
The simple answer is that the smaller players are willing to accept a lower profit margin than the large players are. Whether that rises to the level of gouging or not is a different conversation, but when the smaller players tell you that they don't worry about transit (backhaul) costs when they accept thinner margins than tier 1, it's likely the truth. Or they're lying and will be driven out of business by the backhaul costs. Time will tell.
There's also the consideration of rising utilization and future projected costs, and efforts taken to control the risk that 2020's profits will be negative because everyone is streaming HD4K to 12 TVs in their house.
Cutting back prices and delivering more services cuts into their profit. If we project this across all services, dropping Comcast's profits from 10.5% to 0.5% would mean their $80/month BLAST internet would cost $72/month, and their $300 mega-package would cost $270. Verizon's $50/month packages could become $46.
The other way this works is we find a way to run Verizon's network with 31,000 employees instead of 38,000, and then we do a 7,000-employee layoff. Then Verizon can supply the same services 18% cheaper ($9/month savings!), at the cost of temporarily ticking unemployment up by 0.0047% (4.9% becomes 4.9047%). The savings to the end user would go toward that $9/month Netflix account, meaning Netflix must expand their operations by... well, 7,000 employees, if you actually bought that much labor's worth of services... creating new jobs. (Okay, Netflix, Spotify, some retail--some of the unemployed network technicians will stay unemployed while we buy Chinese imports, necessitating that an already-unemployed, uneducated retail monkey gets a shiny new job at K-Mart.)
That's called technical progress, and it's why services get cheaper and why more complex goods and services become affordable. It's also why we went from hunting, growing, *and* buying food in the 1800s to spending 43% of our income on food in 1900, then 30% in 1950, and 11% today: we've replaced lots and lots and *lots* of farm labor (90% of labor in America in 1790) with very little farm labor (under 2% today) plus substantially more labor building tractors and synthesizing fertilizer for the farmers. The (enormous) set of unaccounted lost labor has gone on to operate retail businesses, fast food drive thrus, shipping companies, warehouses, IT cloud infrastructure services, Netflix, CableTV, ISPs, accounting firms, Amazon, and so forth.
If the technical progress of supplying additional bandwidth utilization per person doesn't remove labor requirements as quickly as people find new ways to use additional bandwidth, then the actual cost of operating these networks will increase. That means Comcast, Verizon, and Level 3 will have to hire more employees for the same number of accounts, pay more for hardware for the same number of accounts ("more hardware" isn't an issue; "an amount of hardware that requires more total cost to provide, thus to purchase" is), pay more total cost laying fiber to support the same number of accounts, etc. Then that $80/month service doesn't cost Comcast $72, but $90; they'll have to charge you $90 to break even, and $100 to keep their 11% profit going.
The other side to this story is we see both bandwidths and data caps in
I use Chrome because it actually uses *less* RAM than Firefox. This is hard to count, because a lot of the RSS is shared memory: the Chrome executable and all loaded libraries are read-only and mapped to the same physical RAM across processes; any writable library data is initially mapped the same way, but gets copied when written (it's read-only but exposed read-write to the program; on protection fault writing, the kernel copies the page to another area of physical RAM, remaps to that, and tells the program to continue); and the process post-fork() does the same with all of its existing data (any aligned 4KB span of data that's initialized before fork() and then unchanged after fork() takes up zero physical RAM in the child process).
The only effective way to count RAM usage in this context is to measure RAM, kill the process tree, and measure RAM again. As-is, all my memory measurement tools tell me Chrome is using more physical RAM than all allocated RAM in my entire system (including swap).
Midbrain is slower than reflex, yes; but it doesn't involve the part of my brain that needs to think and decide. Your entire argument is the prefrontal cortex shuts down and the midbrain goes, "MASH PEDAL HARDER!" The midbrain is capable of complex actions like smoking, scratching your balls, picking at hairs, opening (and finding) bags of food and proceeding to eat, and so forth. When under stress, your midbrain--the part that kicks in and does things when panicking--will start to do these things without alerting you.
This is why martial artists train so damn much. When you lung at someone with a knife, they automatically move forward, avoid the knife, block your arm, punch you in the chin to deflect the rest of your body (it follows the head to avoid neck damage), duck under your arm, put your arm in a lock, and knock you down to your knees. That's not a lot of thinking and planning; that's the brain panicking, assessing the inputs, and applying something it's done thousands of times in similar situations. It might be a long and complex chain of actions requiring the continuous response to sensory input, and the brain can handle that just fine without you being aware of anything more than "oh shit guy is trying to kill me with knife!"
Again: My entire sympathetic nervous system, my hindbrain, and my midbrain have been exposed for 15 years to the various sensory inputs of driving. The 4 inch elevation difference between brakes and gas, the 1 gram accelerator versus the 250 gram brake, the feeling of the car moving forward or backward versus stopping, the visual input of things beyond the windshield, the feeling of torque on the steering wheel when the car moves, all of it. These parts of my brain and even farther down (spinal column...) reflex to the car taking an inappropriate action by making an appropriate response. Granted, I won't automatically grab the handbrake if the foot brake fails; my entire driving experience has primed my system to move onto the footbrake if braking is desired, even if I've mistakenly stomped on the wrong pedal and gotten inappropriate feedback.
Without this, I would constantly crash into things.
LOL now I know you are trolling. You provided that second link with no context whatsoever as to what it is.
You were commenting on the complexity of implementing networks; I showed a distributed network.
the cost of the uplink was the minority
The uplink to what?
The answer: the uplink to the big, complex, EXPENSIVE, distributed back-end network run by the major Tier-1 providers.
You're trying to claim this is just an uplink, a cable you plug into somewhere. What it is is thousands of miles of fiber run between regions, into regional hubs which distribute out to other regional hubs, which talk to each other so a break in connection between Region 1 Hub 13 and Region 1 Central can route to Region 1 Central via Region 1 Hub 7 (and so Region 1 Central has 6 paths to Region 2, or Region 3, or so forth). The last mile is the last mile of fiber run to a little cable on a pole trailing all the way to a little box in a closet at a hub data center, which is part of a sizable data center, which has thousands of miles of fiber uplinking it to other, bigger, more-complex data centers.
The problem is your experience is sitting on the last mile. You're looking down at the run to the end user and saying, "Damn, that's a lot more than the three-foot run to the uplink behind me. It's way expensive. 99% of our cost goes to that last mile." Then you're turning around, looking at the uplink, and going, "That's so cheap. Upstream providers have so little to do." You aren't looking at the upstream provider's network, which is a little three-foot cable running down to you and, on the other end, a massive, distributed, highly-complex network of large data centers spanning the entire fucking country.
How much do you think it costs to expand that back-end with new fiber, new equipment, and new capacity when demand goes up?
The expensive part of the internet connection is the last mile. Each endpoint has a cost to it
So you're saying putting 300,000 computers on the Internet in Comcast's network requires linking exactly 300,000 endpoints? There's no additional cost per endpoint? No back-end costs, no cross-state lines connecting California to Virginia, nothing but a little line run from your house to Comcast's little shack in your cozy little neighborhood?
You're saying the cost of running a ginormous, distributed network with multiple routes to reach from any point to any other point, spanned across an entire continent, is exactly the same (or less!) per customer endpoint as the cost of plugging one customer endpoint *into* that network?
There are approximately 2 billion computers in operation on the planet Earth today.
If you utilized the full processing power of all 2 billion of those computers, you could count from 1 to 2^128 in 1,790,000,000,000 years or 130 times the age of the universe, assuming single-CPU operation. With an average of 6 execution units (6 core) per computer, the iteration can be completed in 298,000,000,000 years or 22 times the current age of the universe.
That's strictly iterating at maximum speed; this excludes the time to make network tests or to compute local hashes or encryptions in an attempt to crack passwords or otherwise-obscured information. Such actions require several thousands to several billions times the amount of real-time.
What if the Tier-3 providers provide the last mile at a certain cost, and imagine it costs Comcast the same or less to provide the giant back-end? I.e. for me to run the cable to your house from the CO 3 miles up the road costs what amortizes out to $12/month; but Comcast is charging $80/month! Maybe I decide Comcast's doing nothing more than what I'm doing, and is charging an assload for it, and passing the cost down to me to pass down to my customers.
What if the Tier-3 providers are positioned to profit by convincing people that the Tier-1 providers are bad and overcharging? What if they perceive Comcast and Verizon as overcharging them, and think they can get their costs down by putting public pressure? What if they think people will just have a bad taste about signing up for a Verizon-branded service and will instead sign up for the little guy's Verizon-proxy?
"Economy of Scale" is a semi-myth. In the general myth, it is observed and true that producing more than minimum quantity of a good uses an optimal amount of per-unit labor; and it is not recognized (by most) that scaling beyond a certain production level increases the amount of per-unit labor (for example: scaling food production up beyond available fertile land requires more fertilizer, pesticide, and irrigation input; increasing technology to farm more densely reduces land per unit output requirement and allows more-efficient use of suboptimal land, allowing further scaling).
Your usage is the less-invoked myth. The last-mile player only has to supply the last mile; Comcast has to supply an enormous back-end with many peering contracts, many interconnects, and much more hardware.
Think about it this way: on my end, I only need an $80 cable modem replaced every 5 years and 20 cents of electricity per month. There is no way Comcast should have to charge me more than $1.50/month for 2Gbit Internet, right? By economies of scale, they should be *far* more effective at networking than I am, since my little cable modem could supply half the block with Internet access if I used one with multiple lines, consuming less electricity per line and costing less than $80 x $LINES for the modem.
Of course Comcast's network has to do a lot more than get the last mile down.
Comcast, who built the Cable network? Comcast who got into a ridiculous legal battle with Level-3 because Level-3 wanted to peer with Comcast to route traffic to parts of the Interent which required Level-3 to cross Comcast's backbone, but Comcast tried (successfully) to make Level-3 pay for peering? Comcast, who supplies Comcast for Business, placing Internet Web sites for independent businesses *directly* on Comcast's network, which is nation-wide, run on Comcast's own equipment, across Comcast's own fiber, and addressed with public IPs, meaning many Web sites simply aren't reachable without going directly through Comcast?
Verizon, AT&T, and Sprint, who built the phone network? You know, the networks which, if they go down, break your ability to reach random Web sites even though you're on some other network, simply because their network is between your ISP and the Web host's ISP?
These are the people Amazon, Netflix, Pandora, Google, Facebook, and Akamai directly establish peering with. These are the people EC2 and Microsoft Azure hook up to so their data centers (you know, S3 and Microsoft Azure Cloud) don't go down off the entire Internet just because Comcast or Verizon or AT&T unplugged the wrong router today.
The people yammering about how much it costs to supply bandwidth are last-mile providers who lease their data lines to the Internet from Comcast, Verizon, AT&T, Sprint, Level 3, XO, CenturyLink, Cox, or others.
Amusing that today's headline is liars accusing other liars of being liars.
These small cable operators don't build the back-end infrastructure of the Internet. They're last-mile carriers. They're like MVNOs who only piggy back on providers, and they're talking about what it's like to build and maintain the Internet backbone as if they actually do that.
So we have these tiny, no-name providers dangled off Verizon, Comcast, and Level 3 talking about the business internals of running Verizon, Comcast, and Level 3?
Maybe next MVNOs who use Verizon and T-Mobile's networks instead of building their own capacity can tell us how very little it costs to maintain national cell networks and admit that Verizon and T-Mobile are just overcharging.
128 bits when all I have to do to find out whether I have the right 128 bits is to send a request with those 128 bits (potentially base64 encoded to get them transferred) and get a response, these 128 bits are rather trivial to crack.
If you use a 3GHz CPU to INC from 0 to 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF (128 bits) at 1 cycle per INC, 3 billion increments per second, directly in register memory, it would take 3,600,000,000,000,000,000,000 years to count. The universe is 13,772,000,000 years old. That's 260,000,000,000 times the current age of the universe--19 times the square of the age of the universe.
It sounds like you are speculating about how you imagine it would be, rather than speaking from experience
No, this is how I actually drive. I *often* have no idea what the hell's going on until after I've done something. I frequently grab hot objects without burning my hand because I react to handling a 700 degree piece of metal by removing my hand from it. On the highway, I usually *predict* what's happening around me; but, when surprised by something I hadn't predicted at all, I immediately respond--often wrongly, by suddenly braking, because braking is a known response; I've done lane tosses frequently enough (instead of braking, you accelerate into an opening to go around an obstacle, which is safer), but not with perfect reliability.
I have *also* slipped my foot off the brake and floored my accelerator while behind a parked car. I didn't hit the car. The first time I did this, I spent quite a while considering htf something like that happens.
This caused the engine to race and the car to suddenly accelerate backwards (since the brake pedal's travel was larger than the accelerator's), and my automatic (muscle-memory) reaction to correct this was to press down on the brake harder, which only caused the car to move faster.
See, when the acceleration kicks in, my brain goes, "Oh you must be on the accelerator!" The brainstem picks up on that as the midbrain feeds it sensory information (sound, visual cues, the feeling of my body as the car moves in an unexpected direction) faster than it gets to my prefrontal cortex. It then issues commands before I have a chance to assess and react to the situation. My body does the right thing.
"Having part of one's brain removed" is a pretty good description of what happens when a person panics -- the intellectual part of their brain shuts down and the low-level flight-or-fight subsystem takes over.
This has only happened to me during a psychotic episode. Particularly, it's only happened when I start hallucinating severely at night, alone, in the dark, due to nyctophobia triggered by being alone, in the dark, at night. My visual cortex projects an internal visualization of the current visual field and then alters it; the rest of my brain responds to the altered version, as it coincides with the original aside from a few projected specters. I'm capable of walking around anywhere blindfolded because of this.
In any case, my brain responds to stimulus with an appropriate reaction based on very blunt reasoning. My response-inhibition system requires the involvement of my prefrontal cortex; but "I want X and action A will get X" is what you'd get if you removed my entire frontal lobe, and it's enough for my brain to reason that the feeling of movement is incorrect and can be corrected by turning the steering wheel or pressing the brake or whatever is required.
and there may not be any feeling of spooling up torque
This is the feeling of the car--or, more precisely, the seat your ass and back are in (or suddenly applying force into)--moving and applying force against you. If there is no feeling of torque kicking in on the wheels, there is no acceleration.
The pedal is still softer. Accelerators can generally go down with 1g-5g; and modern one-pedal driving cars apply resistive braking when you let off the pedal, so braking would kick in immediately as the driver moved to the brake in response to touching a really soft pedal and getting kicked into her seat by insane amounts of instant torque.
There was a hole in the front of my gym from someone doing it this last winter, and it was the second time that person had done it!
Look, I don't sit around when something happens and go, "Huh? What's going on? Wait, I know this feeling. Why is this happen now? This shouldn't be happening now! OH SHIT!" If I did, I'd have third-degree burns from repeatedly grabbing the handles of smoking-hot cast iron frying pans (I don't learn very well from negative consequences; although psychologists claim positive punishment doesn't extinguish behavior, and so the avoidance of burns by immediately withdrawing my hand when I touch a 700-degree iron handle is an effective a way of removing the negative consequence while also frequently allowing me to grab an only-warm pan, so maybe I'm just optimized to not waste time fetching a pot holder until I discover a hot handle).
I assume other people reflexively respond to familiar and unwanted stimulus by taking mitigating actions they've learned through thousands of repetitions. You've hit the accelerator hundreds of thousands of times, and encountered the perception of movement; you've hit the brake thousands of times; if you hit the accelerator, you should recognize that, and know how to undo it.
To be fair, I did fail to handle a new situation recently. A cat ran across a highway in front of me. Going 50mph and not having registered that there were no cars behind me, I used the normal strategy of partial braking while checking my mirrors; I did not stop fast enough. Maximum braking force would have inflicted minor injury on me and my passenger (my car can stop with ~10-12 linear Gs; it's like hitting a tree), but nothing serious. It took me between 2 and 3 real-time seconds to assess the situation (the subjective memory was *considerably* distorted and projected ~10-15 seconds; this required correction), meaning I was physically and mentally capable of stopping in around 1/3 that distance *if* my brain would have selected for maximum braking force with the same reaction speed.
I've replayed that event several hundred times since, and removed all the psychiatric trauma; the new responses are only weakly embedded, so good luck if your kid walks out from behind a parked van 4 feet in front of me in a 30mph speed zone. My normal manner of driving involves scanning the area ahead, including sidewalks, to account for pedestrians and other independent behaviors; the cat exercised an unpredicted behavior before I recognized there was a cat, so that situation should not frequently translate to unleashed humans. I think.
None of this would matter if my brain couldn't select from a set of known behaviors based on a set of recognized conditions *before* informing me wtf is going on and why I'm doing something. Without that, I'd just crash into things all the time. I assumed everyone else had similar facilities, since it's a core part of human neurology.
How do you identify the unique ID of the message? If the message is sent to you (or a group including you), I guess that works. How else?
If message unique IDs are cryptographically secure--if they're 128-bit random GUIDs from a strong entropy source--then this is like saying an attacker only needs the unique private key to hijack Verisign. If they're akin to the ObjectID in MongoDB--datestamp, machine, process, and 24-bit random counter--then we can go fishing. If the ID is discoverable only by being the logged-in user, then you need a browser-end hijack or a TLS-breaking MITM, in which case there are any number of ways to invisibly send messages and not send messages the user types.
Uber is the only one who complained; and I am suggesting a consequence of marketing when making the best effort to supply a service within the declared and current business model of an existing company. I never suggested a business model.
It is impossible for a potentiometer to fail in that manner intermittently. The potentiometer would have that failure in that position 100% of the time.
Brakes are generally heavy and require much more force than accelerators. Clutches require much more travel distance. If you did, somehow, miss the brake and hit the accelerator, you'd probably hit the accelerator hard enough to floor it--or just enough to nudge the brake gently to engaging.
I brake all the time while the clutch is engaged. As long as I don't go under ~500RPM, the car doesn't stall; dropping some speed by braking while in gear won't stall the engine.
Making the connection that the "brake pedal" is actually causing acceleration and switching to the proper pedal is a higher-level brain function that happens much more slowly.
It happens before my car starts moving. If I hear the engine start to growl or feel any form of torque upon depressing a pedal, I know what just happened. The accelerator is also a much lower-force pedal than the brake. There is so much instant feedback telling me what I just did I'd have to have part of my brain physically removed to not know wtf just happened. I'm pretty sure my autonomous nervous system reacts before my prefrontal cortex processes the information, too.
Slashdot Mirror
It's simple enough for me. I just looked at my system this weekend and realized I'm not even running LVM, and that it's a holy hell of stuff that just kind of fell into place because I haven't been picking it apart with tweezers. I've been running as a regular old user, using the software center and GUI tools for a long time; I even used the Software Center to install Steam by clicking the "Software" icon and typing "Steam".
At the same time, I'm designing a new swap daemon to manage swap on Linux. That whole swap partition thing is dumb now, because the installer on various distributions (notably Debian-ites) frequently allocates hundreds of gigabytes of swap when you have 64 or 96 gigabytes of RAM. The right solution would be an automatic swap manager using various strategies; my primary strategy is a zram swap strategy (allocate more zram devices, coalesce them, and otherwise increase zram as memory pressure increases until a fixed limit--say half of RAM is holding compressed data), and other workloads may need alternates such as zswap (uses swap files, and caches them in compressed memory--less than 100% of swap is in RAM), swap files, or LVM swap partitions created out of free volume group space on-the-fly.
A swap daemon like that would free the user from worrying about what the installer automatically does (e.g. eat 110GB of your 128GB SSD for a single swap partition), how much swap space they need up-front, and so forth. The extended configuration allows reconfiguration for specialized workloads (notably, scientific computing). In normal function, the computer would just behave better by default, OOMing less, struggling less when memory usage is high, and not consuming disk space for memory swap space; the user wouldn't even know about the new swap daemon unless someone told him.
That's what all this complex esoterism is: a bunch of shit that goes under the hood. Nerds care about it and we talk about it; end users only know that their graphics card actually works in this release.
It's out of the software's control, and still is completely relevant. The strategy of relying on shared pages makes the difference between whether a multiple-process model is *necessarily* huge bloat or easily-optimized to not bloat. For the application developer, it means designing the application to make some runtime state more stable (i.e. it's different per-run, but less-likely to change during a run) or to group strongly-stable state together (e.g. mmap() anonymous memory for blocks of things which repeat down to the child processes but don't change after they're allocated in the parent) reduces total physical memory usage, even though this may increase measured memory usage.
The software can't make it happen or detect it happening; it can account for it being likely.
Everyone does not struggle. There are reasons you have difficulty--not that you're too stupid for college, or that the next guy has a better brain; it's that you're using the wrong methods, and you're entering an unfamiliar environment.
The brain, first and foremost, is an energy-hungry organ. To minimize energy usage, it restructures to readily follow the most common set of actions. Overriding this--self-activation or response-inhibition--requires first formulating a plan in the prefrontal cortex, then activating the dorsolateral prefrontal cortex to directly control the midbrain. In short: your brain out of high school is a high school brain, and making it take actions different in *any* way from high school requires using one part of your brain to force the rest of it into compliance. Those parts of your brain all consume energy and, just as doing an endless stream of push-ups will eventually leave your muscles physically-incapable of lifting any more, your brain will run down the ATP supply in involved cells and become fatigued and, eventually, physically-incapable of taking these actions.
Many people work around this by setting schedules and reframing.
Setting schedules creates a low-energy activation system: your brain, over an average six weeks of effort, restructures such that keeping to that scheduled set of actions is a lower-energy behavior. Over years and with the experience of repeatedly changing schedules as necessary, your brain even develops the facility to set new schedules and adhere strongly to them in a few weeks as *the* lowest-energy behavior. Lowering the energy expenditure makes it easier to cope with something like a move from high school to college.
Reframing reduces energy expenditure by connecting one impulse to another. If you can evaluate an activity in terms of how it supports something you are actively interested in, that activity becomes a low-energy pursuit. Think of it like changing your work into a distraction: studying engineering is hard and boring, but you *like* tinkering with engines and building go karts and such; studying engineering will let you build *better* go karts, and that connection draws your brain to react on impulse by studying engineering. When that study converts quickly and recognizably into a greater depth of enjoyment of your hobby, you get positive reinforcement, causing your brain to structure itself around seeking those results by executing those actions.
Besides all of this, the geniuses among us are only using competent mental techniques. These range from scheduling and reframing as above to structured study techniques (SQW4R, OK4R), casual or systemic mnemonics (simple visualization up through mind palaces and other systems), and even structured note-taking systems (Cornell notes are the best generic; there are better systems for certain specific materials). Developing and exercising the correct skills *vastly* reduces the difficulties faced in college because time spent studying is optimized: less effort and less time produces more complete understanding and better grades.
Telling people to grin and bear it because college is hard is idiotic. If the effort and cost of college are high enough, it's *not* worth the investment--I dropped out and it was the best decision of my life; this may have been wholly untrue if I had employed more-advanced study techniques (and, in fact, I've gone back and structured my own computer science program to self-teach on the weekends because it's a six-month task if I can find 1 hour per day). Identify *why* they're struggling and fix it.
The built-in browser tools know which virtual pages are mapped to the same physical pages, even though knowing that information would require getting kernel-level access to read the page table entries for the process? Does Chrome install a kernel module?
The amount of RAM used by a set of processes which takes the exact same actions (down to the areas of memory allocated and the exact memory addresses used for every byte of data) will be different between Linux and Windows.
The simple answer is that the smaller players are willing to accept a lower profit margin than the large players are. Whether that rises to the level of gouging or not is a different conversation, but when the smaller players tell you that they don't worry about transit (backhaul) costs when they accept thinner margins than tier 1, it's likely the truth. Or they're lying and will be driven out of business by the backhaul costs. Time will tell.
There's also the consideration of rising utilization and future projected costs, and efforts taken to control the risk that 2020's profits will be negative because everyone is streaming HD4K to 12 TVs in their house.
There's a dialogue here with people asserting that bandwidth is *way* cheap and suggesting that Comcast, Verizon, and co are making enormous profits. Comcast's rising profit margins are up around 11% and averaging under 10.5% over the past several years, staying positive; AT&T frequently operates in negative, ranging from 15% losses to 20% gains, averaging around 8% profit; Verizon is similar to AT&T.
Cutting back prices and delivering more services cuts into their profit. If we project this across all services, dropping Comcast's profits from 10.5% to 0.5% would mean their $80/month BLAST internet would cost $72/month, and their $300 mega-package would cost $270. Verizon's $50/month packages could become $46.
The other way this works is we find a way to run Verizon's network with 31,000 employees instead of 38,000, and then we do a 7,000-employee layoff. Then Verizon can supply the same services 18% cheaper ($9/month savings!), at the cost of temporarily ticking unemployment up by 0.0047% (4.9% becomes 4.9047%). The savings to the end user would go toward that $9/month Netflix account, meaning Netflix must expand their operations by... well, 7,000 employees, if you actually bought that much labor's worth of services... creating new jobs. (Okay, Netflix, Spotify, some retail--some of the unemployed network technicians will stay unemployed while we buy Chinese imports, necessitating that an already-unemployed, uneducated retail monkey gets a shiny new job at K-Mart.)
That's called technical progress, and it's why services get cheaper and why more complex goods and services become affordable. It's also why we went from hunting, growing, *and* buying food in the 1800s to spending 43% of our income on food in 1900, then 30% in 1950, and 11% today: we've replaced lots and lots and *lots* of farm labor (90% of labor in America in 1790) with very little farm labor (under 2% today) plus substantially more labor building tractors and synthesizing fertilizer for the farmers. The (enormous) set of unaccounted lost labor has gone on to operate retail businesses, fast food drive thrus, shipping companies, warehouses, IT cloud infrastructure services, Netflix, CableTV, ISPs, accounting firms, Amazon, and so forth.
If the technical progress of supplying additional bandwidth utilization per person doesn't remove labor requirements as quickly as people find new ways to use additional bandwidth, then the actual cost of operating these networks will increase. That means Comcast, Verizon, and Level 3 will have to hire more employees for the same number of accounts, pay more for hardware for the same number of accounts ("more hardware" isn't an issue; "an amount of hardware that requires more total cost to provide, thus to purchase" is), pay more total cost laying fiber to support the same number of accounts, etc. Then that $80/month service doesn't cost Comcast $72, but $90; they'll have to charge you $90 to break even, and $100 to keep their 11% profit going.
The other side to this story is we see both bandwidths and data caps in
It's more ridiculous than that.
I use Chrome because it actually uses *less* RAM than Firefox. This is hard to count, because a lot of the RSS is shared memory: the Chrome executable and all loaded libraries are read-only and mapped to the same physical RAM across processes; any writable library data is initially mapped the same way, but gets copied when written (it's read-only but exposed read-write to the program; on protection fault writing, the kernel copies the page to another area of physical RAM, remaps to that, and tells the program to continue); and the process post-fork() does the same with all of its existing data (any aligned 4KB span of data that's initialized before fork() and then unchanged after fork() takes up zero physical RAM in the child process).
The only effective way to count RAM usage in this context is to measure RAM, kill the process tree, and measure RAM again. As-is, all my memory measurement tools tell me Chrome is using more physical RAM than all allocated RAM in my entire system (including swap).
Midbrain is slower than reflex, yes; but it doesn't involve the part of my brain that needs to think and decide. Your entire argument is the prefrontal cortex shuts down and the midbrain goes, "MASH PEDAL HARDER!" The midbrain is capable of complex actions like smoking, scratching your balls, picking at hairs, opening (and finding) bags of food and proceeding to eat, and so forth. When under stress, your midbrain--the part that kicks in and does things when panicking--will start to do these things without alerting you.
This is why martial artists train so damn much. When you lung at someone with a knife, they automatically move forward, avoid the knife, block your arm, punch you in the chin to deflect the rest of your body (it follows the head to avoid neck damage), duck under your arm, put your arm in a lock, and knock you down to your knees. That's not a lot of thinking and planning; that's the brain panicking, assessing the inputs, and applying something it's done thousands of times in similar situations. It might be a long and complex chain of actions requiring the continuous response to sensory input, and the brain can handle that just fine without you being aware of anything more than "oh shit guy is trying to kill me with knife!"
Again: My entire sympathetic nervous system, my hindbrain, and my midbrain have been exposed for 15 years to the various sensory inputs of driving. The 4 inch elevation difference between brakes and gas, the 1 gram accelerator versus the 250 gram brake, the feeling of the car moving forward or backward versus stopping, the visual input of things beyond the windshield, the feeling of torque on the steering wheel when the car moves, all of it. These parts of my brain and even farther down (spinal column...) reflex to the car taking an inappropriate action by making an appropriate response. Granted, I won't automatically grab the handbrake if the foot brake fails; my entire driving experience has primed my system to move onto the footbrake if braking is desired, even if I've mistakenly stomped on the wrong pedal and gotten inappropriate feedback.
Without this, I would constantly crash into things.
LOL now I know you are trolling. You provided that second link with no context whatsoever as to what it is.
You were commenting on the complexity of implementing networks; I showed a distributed network.
the cost of the uplink was the minority
The uplink to what?
The answer: the uplink to the big, complex, EXPENSIVE, distributed back-end network run by the major Tier-1 providers.
You're trying to claim this is just an uplink, a cable you plug into somewhere. What it is is thousands of miles of fiber run between regions, into regional hubs which distribute out to other regional hubs, which talk to each other so a break in connection between Region 1 Hub 13 and Region 1 Central can route to Region 1 Central via Region 1 Hub 7 (and so Region 1 Central has 6 paths to Region 2, or Region 3, or so forth). The last mile is the last mile of fiber run to a little cable on a pole trailing all the way to a little box in a closet at a hub data center, which is part of a sizable data center, which has thousands of miles of fiber uplinking it to other, bigger, more-complex data centers.
The problem is your experience is sitting on the last mile. You're looking down at the run to the end user and saying, "Damn, that's a lot more than the three-foot run to the uplink behind me. It's way expensive. 99% of our cost goes to that last mile." Then you're turning around, looking at the uplink, and going, "That's so cheap. Upstream providers have so little to do." You aren't looking at the upstream provider's network, which is a little three-foot cable running down to you and, on the other end, a massive, distributed, highly-complex network of large data centers spanning the entire fucking country.
How much do you think it costs to expand that back-end with new fiber, new equipment, and new capacity when demand goes up?
The expensive part of the internet connection is the last mile. Each endpoint has a cost to it
So you're saying putting 300,000 computers on the Internet in Comcast's network requires linking exactly 300,000 endpoints? There's no additional cost per endpoint? No back-end costs, no cross-state lines connecting California to Virginia, nothing but a little line run from your house to Comcast's little shack in your cozy little neighborhood?
You're saying the cost of running a ginormous, distributed network with multiple routes to reach from any point to any other point, spanned across an entire continent, is exactly the same (or less!) per customer endpoint as the cost of plugging one customer endpoint *into* that network?
You're saying this is the expensive part and this is the cheap part?
You're using the argument: "Y must be more expensive than X because Y includes X + P".
What if P is much less than X, and provider of Y claims X is *exactly* the same as P, and thus X is wildly overpriced?
What if P is running the last mile, and X is the sum total of all operations a Tier-1 carries out to maintain and expand their network?
There are approximately 2 billion computers in operation on the planet Earth today.
If you utilized the full processing power of all 2 billion of those computers, you could count from 1 to 2^128 in 1,790,000,000,000 years or 130 times the age of the universe, assuming single-CPU operation. With an average of 6 execution units (6 core) per computer, the iteration can be completed in 298,000,000,000 years or 22 times the current age of the universe.
That's strictly iterating at maximum speed; this excludes the time to make network tests or to compute local hashes or encryptions in an attempt to crack passwords or otherwise-obscured information. Such actions require several thousands to several billions times the amount of real-time.
What if the Tier-3 providers provide the last mile at a certain cost, and imagine it costs Comcast the same or less to provide the giant back-end? I.e. for me to run the cable to your house from the CO 3 miles up the road costs what amortizes out to $12/month; but Comcast is charging $80/month! Maybe I decide Comcast's doing nothing more than what I'm doing, and is charging an assload for it, and passing the cost down to me to pass down to my customers.
What if the Tier-3 providers are positioned to profit by convincing people that the Tier-1 providers are bad and overcharging? What if they perceive Comcast and Verizon as overcharging them, and think they can get their costs down by putting public pressure? What if they think people will just have a bad taste about signing up for a Verizon-branded service and will instead sign up for the little guy's Verizon-proxy?
"Economy of Scale" is a semi-myth. In the general myth, it is observed and true that producing more than minimum quantity of a good uses an optimal amount of per-unit labor; and it is not recognized (by most) that scaling beyond a certain production level increases the amount of per-unit labor (for example: scaling food production up beyond available fertile land requires more fertilizer, pesticide, and irrigation input; increasing technology to farm more densely reduces land per unit output requirement and allows more-efficient use of suboptimal land, allowing further scaling).
Your usage is the less-invoked myth. The last-mile player only has to supply the last mile; Comcast has to supply an enormous back-end with many peering contracts, many interconnects, and much more hardware.
Think about it this way: on my end, I only need an $80 cable modem replaced every 5 years and 20 cents of electricity per month. There is no way Comcast should have to charge me more than $1.50/month for 2Gbit Internet, right? By economies of scale, they should be *far* more effective at networking than I am, since my little cable modem could supply half the block with Internet access if I used one with multiple lines, consuming less electricity per line and costing less than $80 x $LINES for the modem.
Of course Comcast's network has to do a lot more than get the last mile down.
You mean Verizon, Comcast, AT&T, Sprint, etc.?
Comcast, who built the Cable network? Comcast who got into a ridiculous legal battle with Level-3 because Level-3 wanted to peer with Comcast to route traffic to parts of the Interent which required Level-3 to cross Comcast's backbone, but Comcast tried (successfully) to make Level-3 pay for peering? Comcast, who supplies Comcast for Business, placing Internet Web sites for independent businesses *directly* on Comcast's network, which is nation-wide, run on Comcast's own equipment, across Comcast's own fiber, and addressed with public IPs, meaning many Web sites simply aren't reachable without going directly through Comcast?
Verizon, AT&T, and Sprint, who built the phone network? You know, the networks which, if they go down, break your ability to reach random Web sites even though you're on some other network, simply because their network is between your ISP and the Web host's ISP?
These are the people Amazon, Netflix, Pandora, Google, Facebook, and Akamai directly establish peering with. These are the people EC2 and Microsoft Azure hook up to so their data centers (you know, S3 and Microsoft Azure Cloud) don't go down off the entire Internet just because Comcast or Verizon or AT&T unplugged the wrong router today.
The people yammering about how much it costs to supply bandwidth are last-mile providers who lease their data lines to the Internet from Comcast, Verizon, AT&T, Sprint, Level 3, XO, CenturyLink, Cox, or others.
Amusing that today's headline is liars accusing other liars of being liars.
These small cable operators don't build the back-end infrastructure of the Internet. They're last-mile carriers. They're like MVNOs who only piggy back on providers, and they're talking about what it's like to build and maintain the Internet backbone as if they actually do that.
So we have these tiny, no-name providers dangled off Verizon, Comcast, and Level 3 talking about the business internals of running Verizon, Comcast, and Level 3?
Maybe next MVNOs who use Verizon and T-Mobile's networks instead of building their own capacity can tell us how very little it costs to maintain national cell networks and admit that Verizon and T-Mobile are just overcharging.
128 bits when all I have to do to find out whether I have the right 128 bits is to send a request with those 128 bits (potentially base64 encoded to get them transferred) and get a response, these 128 bits are rather trivial to crack.
If you use a 3GHz CPU to INC from 0 to 0xFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFFF (128 bits) at 1 cycle per INC, 3 billion increments per second, directly in register memory, it would take 3,600,000,000,000,000,000,000 years to count. The universe is 13,772,000,000 years old. That's 260,000,000,000 times the current age of the universe--19 times the square of the age of the universe.
How trivial is trivial?
It sounds like you are speculating about how you imagine it would be, rather than speaking from experience
No, this is how I actually drive. I *often* have no idea what the hell's going on until after I've done something. I frequently grab hot objects without burning my hand because I react to handling a 700 degree piece of metal by removing my hand from it. On the highway, I usually *predict* what's happening around me; but, when surprised by something I hadn't predicted at all, I immediately respond--often wrongly, by suddenly braking, because braking is a known response; I've done lane tosses frequently enough (instead of braking, you accelerate into an opening to go around an obstacle, which is safer), but not with perfect reliability.
I have *also* slipped my foot off the brake and floored my accelerator while behind a parked car. I didn't hit the car. The first time I did this, I spent quite a while considering htf something like that happens.
This caused the engine to race and the car to suddenly accelerate backwards (since the brake pedal's travel was larger than the accelerator's), and my automatic (muscle-memory) reaction to correct this was to press down on the brake harder, which only caused the car to move faster.
See, when the acceleration kicks in, my brain goes, "Oh you must be on the accelerator!" The brainstem picks up on that as the midbrain feeds it sensory information (sound, visual cues, the feeling of my body as the car moves in an unexpected direction) faster than it gets to my prefrontal cortex. It then issues commands before I have a chance to assess and react to the situation. My body does the right thing.
"Having part of one's brain removed" is a pretty good description of what happens when a person panics -- the intellectual part of their brain shuts down and the low-level flight-or-fight subsystem takes over.
This has only happened to me during a psychotic episode. Particularly, it's only happened when I start hallucinating severely at night, alone, in the dark, due to nyctophobia triggered by being alone, in the dark, at night. My visual cortex projects an internal visualization of the current visual field and then alters it; the rest of my brain responds to the altered version, as it coincides with the original aside from a few projected specters. I'm capable of walking around anywhere blindfolded because of this.
In any case, my brain responds to stimulus with an appropriate reaction based on very blunt reasoning. My response-inhibition system requires the involvement of my prefrontal cortex; but "I want X and action A will get X" is what you'd get if you removed my entire frontal lobe, and it's enough for my brain to reason that the feeling of movement is incorrect and can be corrected by turning the steering wheel or pressing the brake or whatever is required.
and there may not be any feeling of spooling up torque
This is the feeling of the car--or, more precisely, the seat your ass and back are in (or suddenly applying force into)--moving and applying force against you. If there is no feeling of torque kicking in on the wheels, there is no acceleration.
The pedal is still softer. Accelerators can generally go down with 1g-5g; and modern one-pedal driving cars apply resistive braking when you let off the pedal, so braking would kick in immediately as the driver moved to the brake in response to touching a really soft pedal and getting kicked into her seat by insane amounts of instant torque.
There was a hole in the front of my gym from someone doing it this last winter, and it was the second time that person had done it!
Look, I don't sit around when something happens and go, "Huh? What's going on? Wait, I know this feeling. Why is this happen now? This shouldn't be happening now! OH SHIT!" If I did, I'd have third-degree burns from repeatedly grabbing the handles of smoking-hot cast iron frying pans (I don't learn very well from negative consequences; although psychologists claim positive punishment doesn't extinguish behavior, and so the avoidance of burns by immediately withdrawing my hand when I touch a 700-degree iron handle is an effective a way of removing the negative consequence while also frequently allowing me to grab an only-warm pan, so maybe I'm just optimized to not waste time fetching a pot holder until I discover a hot handle).
I assume other people reflexively respond to familiar and unwanted stimulus by taking mitigating actions they've learned through thousands of repetitions. You've hit the accelerator hundreds of thousands of times, and encountered the perception of movement; you've hit the brake thousands of times; if you hit the accelerator, you should recognize that, and know how to undo it.
To be fair, I did fail to handle a new situation recently. A cat ran across a highway in front of me. Going 50mph and not having registered that there were no cars behind me, I used the normal strategy of partial braking while checking my mirrors; I did not stop fast enough. Maximum braking force would have inflicted minor injury on me and my passenger (my car can stop with ~10-12 linear Gs; it's like hitting a tree), but nothing serious. It took me between 2 and 3 real-time seconds to assess the situation (the subjective memory was *considerably* distorted and projected ~10-15 seconds; this required correction), meaning I was physically and mentally capable of stopping in around 1/3 that distance *if* my brain would have selected for maximum braking force with the same reaction speed.
I've replayed that event several hundred times since, and removed all the psychiatric trauma; the new responses are only weakly embedded, so good luck if your kid walks out from behind a parked van 4 feet in front of me in a 30mph speed zone. My normal manner of driving involves scanning the area ahead, including sidewalks, to account for pedestrians and other independent behaviors; the cat exercised an unpredicted behavior before I recognized there was a cat, so that situation should not frequently translate to unleashed humans. I think.
None of this would matter if my brain couldn't select from a set of known behaviors based on a set of recognized conditions *before* informing me wtf is going on and why I'm doing something. Without that, I'd just crash into things all the time. I assumed everyone else had similar facilities, since it's a core part of human neurology.
How do you identify the unique ID of the message? If the message is sent to you (or a group including you), I guess that works. How else?
If message unique IDs are cryptographically secure--if they're 128-bit random GUIDs from a strong entropy source--then this is like saying an attacker only needs the unique private key to hijack Verisign. If they're akin to the ObjectID in MongoDB--datestamp, machine, process, and 24-bit random counter--then we can go fishing. If the ID is discoverable only by being the logged-in user, then you need a browser-end hijack or a TLS-breaking MITM, in which case there are any number of ways to invisibly send messages and not send messages the user types.
Uber is the only one who complained; and I am suggesting a consequence of marketing when making the best effort to supply a service within the declared and current business model of an existing company. I never suggested a business model.
It is impossible for a potentiometer to fail in that manner intermittently. The potentiometer would have that failure in that position 100% of the time.
Brakes are generally heavy and require much more force than accelerators. Clutches require much more travel distance. If you did, somehow, miss the brake and hit the accelerator, you'd probably hit the accelerator hard enough to floor it--or just enough to nudge the brake gently to engaging.
I brake all the time while the clutch is engaged. As long as I don't go under ~500RPM, the car doesn't stall; dropping some speed by braking while in gear won't stall the engine.
Making the connection that the "brake pedal" is actually causing acceleration and switching to the proper pedal is a higher-level brain function that happens much more slowly.
It happens before my car starts moving. If I hear the engine start to growl or feel any form of torque upon depressing a pedal, I know what just happened. The accelerator is also a much lower-force pedal than the brake. There is so much instant feedback telling me what I just did I'd have to have part of my brain physically removed to not know wtf just happened. I'm pretty sure my autonomous nervous system reacts before my prefrontal cortex processes the information, too.