Of course. Because Google ads are all over the place, not only on google.com/search. Should they ever decide to get out of one of those two, I'm pretty sure they'd rather get out of the search business than the ad business.
Let's not be intentionally thick here. Obviously they have more than one web service. But between discontinuing all the services vs. discontinuing all the ads and finding some other method to monetize the services, the latter is obviously preferable.
But the only real monetization strategy for virtually all of those "businesses" is ads. Their business is in building interesting stuff that people want to use, so they can get eyeballs on their ads, so they can pay for building interesting stuff that people want to use, so **STACKOVERFLOW**.
That isn't what's in dispute. But you're confusing the monetization strategy with the business. You don't say that The New York Times is in the advertising business, you say that they're in the newspaper business, notwithstanding that most of their revenues come from ads. Likewise, you don't say that Ford is in the banking industry just because they sell most of their cars on credit.
I love knee-jerk responses just as much as the next guy, but please slowly read again what I wrote.
Does Google make any money from letting you search the web for free? No. Does Google make shitloads of money from displaying ads next to those free search results? Hell yeah.
So which one is their business?
OK, I'll bite. Let's determine this by taking each one of them away.
If you take away the ads and put up a paywall, do they still make money? Yes. Probably not as much, because paywalls are bunkum, but some significant number of people would subscribe to Google search and they would still make a huge pile of money.
Now, if you take away the search engine and just have a page with ads on it, do they still make money? Of course not. Nobody is going to keep going back to a page that has only ads on it and nobody is going to pay for ads that nobody is looking at.
Most people never imply that when they say that consumers are the "product". Google is not the only company that does this. There are a ton of other companies out there which provide lead generation services and sell existing databases of leads and demographics.
Except that they don't sell databases or demographic information.
In that context, the consumer is the product. Specifically, the information about them.
Except that both of those things are wrong. If the information is the product then you are not the product because you are not information. Moreover, information about users is not either of the products. Advertising is one, web services are the other.
The average user of Google services is the product, the advertisers are the consumer.
That would imply that, having bought you, the "product," the advertiser now owns you. Which is clearly wrong.
That is a contradiction. Which is it? Customers pay money. That's how it works. Otherwise they are not customers. You have buyers and sellers, products, and end users. You can tell which is which by following the money.
That isn't how it works. If I go to a tag sale and buy a gadget, I am the customer. If I negotiate to trade my widget for the gadget, I am still the gadget customer, I am just not paying in cash. It has nothing to do with paying money.
More importantly, your argument is at best that the relation is reversed. However, if the advertiser is the consumer then you must be the seller, not the product, because in either event Google is the intermediary. In addition to that, it still ignores that there are two products (web services and advertising) and that you are still the customer of the web services, even if you are the seller of advertising (which is still silly -- you aren't producing advertising, you're consuming it).
Your mental and semantic gymnastics aside, suppliers don't typically pay retail stores. That is like Walmart charging Chinese companies and manufacturers for the privilege of selling their product. That is not how it works (kickbacks aside). Walmart pays for the product from the suppliers, and then delivers it to the consumer in their retail stores.
Except that as I already explained, advertising is a product with negative value to the consumer. It's like industrial waste. Nobody wants it, so you have to pay people to take it from you.
In order for me to buy into your post I would need to believe that Google is going to ignore the people giving them the money to favor people that do not. Sorry, but that makes no sense.
Your argument flies in the face of all empirical evidence. Advertisers want huge, obnoxious, intrusive ads. Users want small, polite, unobtrusive ads. Advertisers want web pages that keep users there for as long as possible to increase the probability that you'll see or click the ads. Users want to find what they're looking for quickly so they can move on. The users won.
You're so fixated on who pays money that you're losing sight of who has the economic power. The advertisers don't pay a dime if there are no users to consume ads. Which means you have to do what the users want, because losing users means losing money.
Correction: Google's main business is ads. They just happen to go very well with search.
Can we please put a stop to this FUD? You're just repeating Microsoft talking points.
Google does not own you. The only way they get access to your eyeballs is with your consent. You are the customer, not the product. It is 2012, people are not products. Advertising is a product. You consume the product. It is a product with negative value, so Google pays you to consume it, in the form of barter for services. They must provide you with enough value to keep you consuming the product, because you are the customer, not the product. You are the customer of the ads and the customer of the services. The advertisers are not customers, they are suppliers. They supply one of the products, the ads. Google provides the other product, the services. Google is in the web service industry, because that is the product they actually make, and the one that actually has value to you, the customer. There are no ads without services.
It bears keeping in mind why the exchange of services for ads takes place. They could charge money instead. The reason that they don't is that the typical customer values their own money more than they value the cost of consuming that amount of advertising. You would rather consume an ad than pay a penny. The advertiser is willing to pay a penny for you to consume an ad. In consequence, instead of a straight transaction of web services for money, we have a three-party transaction where you consume Google web services and pay for them by consuming some third party's advertising, because that transaction is Pareto efficient over the one where you pay money for services. But that doesn't make the provider of web services an advertising company. They're still providing web services and you are still choosing whether to consume them over the alternatives.
Now, you might point out, Google owns Doubleclick and operates an advertising network. And in that respect their business is advertising -- but that isn't their main business. If Google discontinued all of their web services but continued their ad network as supplied to third parties, they would lose the large majority of their revenues. If they kept the web services and then discontinued their internal advertising network and instead used a third party's, they would keep the large majority of their revenues. What does that tell you about which one is their main business?
How so? The server has to provide the correct certificate before the client sends the User-agent: header. Otherwise, the server would know which site's cert to use from the Host: header, which is sent just before the User-agent: header.
When the user types e.g. 'gmail.com' the default is the http version, which can do a redirect if it detects a browser that supports it. Also, newer browsers can be made to default to https, which solves it neatly because then new browsers default to https and old browsers keep getting http.
Say someone with an SNI browser shares a link with me via e-mail, IM, social network, or however, but my browser doesn't support SNI. Because the person who sent the link is using an SNI browser, the user was redirected to HTTPS and therefore the link is HTTPS. But when I try to open the link, I get a certificate error: "the URL says example.com but the certificate says webhostingcompany.net".
I imagine the same thing as happens if you send them an Office 2007 document when they have Office 2003, or a PDF when they don't have a PDF reader, etc. They type the error into Google, half the results tell them they need updated software, so they download Chrome and it's problem solved.
Such a download won't work for users who aren't the computer's owner or otherwise the administrator.
Chrome defaults to installing in the user's home directory and doesn't require admin rights. I think there is a version of Firefox that does the same thing.
Nor will it work for devices whose browser is part of an operating system package that only the manufacturer and/or carrier can update, such as any Android phone stuck on Android 2.x.
I'm pretty sure you can get other browsers from the market.
Why would people who own a PC based on a P4 or Atom CPU (they're roughly comparable clock for clock) buy a new operating system when they can install Xubuntu?
I don't think they would do either of those things. It certainly doesn't make sense to buy another operating system -- you can get a faster computer with operating system for less than the retail price of Windows.
But as for Xubuntu (or pick your favorite Linux distribution), I imagine it's the same reason people don't do it on newer computers: Lack of awareness, unfamiliarity with a new environment, third party software support, FUD, etc.
And anyway, the point I was trying to make wasn't that SNI would immediately become deployed in April 2014 but that there was a damn good reason not to deploy it before then. Please allow me to correct myself: "This won't change for at least another two years."
I'm still not seeing what requires the wait. The website knows what the client's browser is. There is nothing stopping it from redirecting all users with compatible browsers to the HTTPS version, today. And providing a message to users of older browsers that their shit is old and busted and click here to download the new shiny, without actually preventing them from using the non-HTTPS version in the meantime.
Windows XP isn't going to magically disappear in 2014 when support ends. The number of low-income individuals who continue to make do with some anachronistic Pentium 4 is really quite large, and those people are not about to go out and spend a few hundred bucks they don't have on a new computer or OS just because Microsoft discontinues support. They probably won't even notice, and will replace their OS when the computer dies from being full of dust.
What's going to happen is that XP users will get a message that says "Your version of Internet Explorer is no longer supported, click here to download Firefox or Chrome." And there isn't any particularly good reason for that to happen in 2014 rather than today or in 2016, other than maybe to wait until XP market share drops past some arbitrary threshold.
1) Why would the ISPs care whether the traffic is encrypted or not? It doesn't cost them anything more. 2) All websites are slowly going to 100% HTTPS. The only reason they ever weren't in the first place was that the CPU overhead was too high back when web servers had 1GHz Pentium III Xeons that cost $2000/core. Today practically every chip you can buy has hardware-level encryption support. (There is also an issue about HTTP being a silly protocol that opens more connections and thus requiring more handshakes than it ought to, but that will be fixed soon as well.) 3) Do you recall what DRM is? That's right, it's encryption. What percentage of internet traffic is Netflix again?
Good luck banning encryption when 85+% of all legitimate traffic is encrypted.
Except that...no, not really. Google wouldn't actually know anything more than your ISP (or the NSA) would know by monitoring your traffic, namely, who your friends are but not what you're sharing with them.
There are ways around that. Probably one of the best is to take one of the lightweight boots-in-3-seconds Linux distributions and run OpenGL games in a VM so that you have a native OpenGL implementation. Or even better, do it the other way, run non-Windows as the host OS and then run 'legacy' games in a Windows VM, because the 'older' games will require less fast hardware and therefore care less about any VM overhead.
Another point here is that if you're successful in making OpenGL the standard again, it's very likely that Microsoft will bring back the native implementation on Windows.
But no PPC support for later browsers will send it to the landfill before that.
--- Eventually we'll be unable to access websites that rely on features in recent versions of flash, java or html5.
You can always put Linux on it. Even the latest Ubuntu runs on PowerPC, which I expect includes an updated Firefox.
The disadvantage is no Flash, but you really shouldn't be running Flash on PowerPC anyway because the latest version has serious security unpatched vulnerabilities. And Flash is slowly disappearing anyway -- your iMac will probably be more useful a couple years from now when Flash is dead than it is now!
A lot of people might like if Steam tried to force publishers to create for an independent platform, but it isn't their job. It isn't even to their benefit, since they seem like they're only creating specifications and not the device itself.
I don't know about that. If they became the new de facto gaming platform, that is a strong position to be in. Especially if the "platform" isn't dependent on any particular vendor (like Microsoft), who may at any time decide that they want to try to eat your lunch.
Anyway, judging from the listed specs, this is not a low-end rig by any measure, meaning that it will already be priced noticeably higher than competitors - so $30 (or whatever it is these days?) for OEM Windows is not likely to make a big difference.
I think this ignores that the goal is to provide consistent specs long-term. The sensible way to do that is to make today's high end the standard, so that it will still have adequate performance going forward, even though five years from now the specs will be met by hardware that sells for a price that makes $35 look significant.
The goal would be to get the entire library of Steam games to run on that console from day 1. It's also easier for the developers if they have, essentially, one platform to target for both PC and console.
My thinking was that you could make Steam the platform, in the same way that Debian doesn't care whether the kernel is Linux or FreeBSD.
The advantage of doing it that way is that you have a sort of "cross-platform platform" -- you say that anything that provides X, Y and Z (e.g. x86-64 quad-core min. 3GHz + GPU of Y minimum spec + Z version of OpenGL + whatever else) is compatible. Then if Apple wants to implement the list of prerequisites on a new Apple TV, or Google wants to implement it on the Google TV, or Canonical wants to make a Steam-compatible media center edition of Ubuntu etc., they can. It lets Valve step in to Microsoft's shoes as the platform maintainer and opens up underlying platform innovation to the world rather than just Microsoft, while still providing minimum standards for developers to target.
Obviously you can't instantly port all of the existing games that way, but there is an easy way around that: You require future games to target the new platform (which Valve can itself implement on Windows, assuming it requires anything not there already), and then put "Windows optional" in the specs, so that if you have Windows then you get all past and future games, and if you don't then you still get all future games and anything anyone wants to port to the new Steam platform in order to get the greater number of sales available on Windows + Everything Else than only on Windows.
My understanding of the story is that essentially this will be a Windows computer with Steam installed.
Is there any inherent reason why it would have to be Windows? If you're going for the console market, i.e. not worried about people needing legacy productivity software, what sense does it make to pay the Microsoft tax? Certainly Sony and Nintendo don't do it, and even Microsoft doesn't run the same operating system on XBOX as they do on PCs.
BSD projects benefit from patches and contributions, both from individuals and corporations.
Exactly, which is why all of MacOS X is BSD licensed. Wait.
You clearly must have missed the part of the GP that says "even being able to take those people don't intentionally contribute."
GPL is perfectly usable in a closed management model when the code is used internally, for example when you provide a service not a software product like google.
Except that if you're concerned about that then you use the AGPL.
Second, it is a political belief, not a fact, that denying access to the close management model is beneficial.
Your opinion regarding the empirical consequences of closed management is rejected on the grounds of it not being a fact.
Your license it for money under an alternative license argument is in conflict with your patches from 3rd parties argument, you can not license code that others own the copyright to - look at the Linux kernel being locked into GPL v2 because all the contributors of patches and new features/functionality can't/won't authorize a switch to GPL v3.
OK, they're in conflict. Which one does that disprove? It's neither of them, isn't it? It's that you get a choice between reincorporating changes from unwilling contributors (without copyright assignment, as with the Linux kernel) or requiring copyright assignment from contributors before incorporating their contributions so that you can license them on different terms while preventing prospective licensees from just taking the code and incorporate it into proprietary software without your consent. You don't get to do either of those with BSD.
Debian packages are not really projects. There are projects which are divided into many smaller packages, like Xorg or Libreoffice, and there are packages which contain many small projects in aggregate, e.g. kdeapps.
That really depends on how you define "projects." It's certainly true that most everything in kdeapps was developed by a common collection of people, but is that really the important dividing line? It seems odd to say that, for example, Konqueror and Kate are not separate 'apps' whereas two pieces of software that provide the same functionality are because they happen to have been written by different people or are more often distributed separately.
Well, I think it's pretty obvious that obesity is unrelated to anorexia. The trouble is that you're not just changing the ideal for people who are anorexic, you're changing the ideal for people who are obese as well.
Again, the proper way to deal with anorexia is by educating people on how to live a healthy lifestyle and providing incentives to do so. For that matter, that is pretty well the proper way to deal with obesity too.
Does that help recalibrate your understanding of where I'm coming from? Even a 12x reduction in size, 3x reduction in associativity, a modest "process" advantage, and fewer access ports couldn't give a 2x speedup. You're never going to get even close to halving L2 latency by halving its size. That needs a far bigger hammer.
I understand that. My thinking was that if Intel could manage 11 cycles at 32nm then AMD could too, if no other way than by adopting a cache design similar to the one Intel uses.
Bringing up the idea that Intel's process technology is superior to Global Foundries' rather explains that though, and it gives me another thought: It may be that the reason for the high latencies is intentional in that AMD designed Bulldozer to hit significantly higher clock speeds than it has and high leakage on GF's process is thermally limiting it (and is likewise a significant contributor to the high power consumption).
If that's the case then I think it bodes well for the design, because it means that if GF (or, failing that, TSMC et al) can improve their process technology to reduce leakage, the chips may clock much higher and both performance and performance-per-watt will improve. It also puts an interesting spin on the subject of the article: "Resonant clock mesh" (whatever that is) ostensibly reduces power consumption and heat. If they're successful in implementing it, the design could end up looking much more attractive.
Actually, there is a major benefit to a processor which is 50% faster, even if it's idle much of the time. Power efficiency. There is increasing focus on total energy consumption, to improve the battery life of portable computers. CPU usage profiles for light loads look like "wake up, do something, go back to sleep". If you have two CPUs which both use the same power, but one is 2x as fast, the faster CPU's time spent in wake is literally half as much as the slower CPU. Now throw power gating into the mix -- when idle, modern CPUs can literally be shut down almost completely, consuming no power. This means the 2x faster CPU is literally 2x as energy efficient, even for light loads, because they both consume the same amount of power while active, but the faster CPU is active for much less time.
That is true, though I'm not sure how large a loss that would be in practice: It has minimal effect on the desktop market and portable computers of the "slower CPU / faster GPU / low watts" variety are the wheelhouse of the Bobcat rather than Bulldozer, which takes care of anyone whose primary concern is battery life. Then AMD can still get anyone who prioritizes purchase price over battery life, as well as anyone who prioritizes battery life and better-than-Bobcat performance over weight (since you can use the money you save buying AMD over Intel to install a bigger battery that more than makes up for the power consumption). All it really excludes them from is those who prioritize performance, battery life and weight over price, which has a pretty high overlap with the market that demands the highest single-thread performance at any price (which they had already lost).
Wait until we have real universal health care.. then the anorexics health becomes everyones business... and thus, justifying laws like the banning of skinny models. Think of this research like a wake-up call.
The biggest problem with this isn't even the "big government" problem. It's that the researchers are empirically wrong about the solution.
Are we not presently in the middle of an obesity epidemic? The problem is not that people want to be thin -- that would probably be a good thing. The problem is that they're doing it wrong, i.e. in an unhealthy way rather than through proper diet and exercise. So how can the solution be to make people not want to be thin anymore? That would just result in more couch potatoes and more heart disease.
The solution is to see to it that the people who are dedicated to looking fit are educated about how to do that, i.e. a combination of sit-ups and peanut butter rather than nervous tension and starvation.
Yeah, that's pretty much it. There is no "cyber war" -- but there is state-backed industrial espionage.
And the problem with the way these people are thinking about it is that MAD is totally wrong. It isn't offense that you need, it's defense. Especially for the US: What "the enemy" is doing is sabotage and misappropriation trade secrets on a massive scale. Even if you can manage to do the same thing to them, your infrastructure and trade secrets are more valuable than theirs. Which means that having the same offensive capability as they do means that you lose.
You can't win unless you can prevent yourself from losing. The way to prevent "cyber war" is to keep sensitive industrial equipment disconnected from the internet, provide incentives for relevant people in industry to implement security best practices, etc.
The idea of creating an arsenal of "cyber weapons" is fairly preposterous. About the only good value I can imagine in having one is that it gives you a good threat assessment for things you need to be able to prevent when designing your defenses. The idea that there is ever going to be some kind of doomsday cyberwar where they launch a "cyber attack" to shut down the power grid or some such Hollywood bullshit is totally inane -- even assuming that the enemy launches such an attack, how are your "cyber weapons" going to do anything in response when your country no longer has a working internet? In other words, you need defense, not offense. And if you have a sufficient defense, the offensive capability is superfluous.
I'm going to go ahead and concede that I don't design microprocessors for a living, and I'll give you the heritage of the various architectures because I'm primarily just reiterating what others have told me and I feel like at this point unless you have some kind of non-public information we're both just speculating about it anyway.
So if you want to be the expert, let's see if we can learn anything here instead of just having a flame war.
You're arguing that Bulldozer uses too many transistors for its performance, and never mind that most of those transistors are wasted on an unfortunate cache architecture because it would be too difficult to change that. At the same time, you seem to think nothing of Intel basically starting over entirely with the L2 between Katmai and Coppermine. So let's see if you can help me understand why that is.
If you look at the cache latencies for Bulldozer, they're terrible. The L2 is 21 cycles compared to 11 for Core i5. The L3 is 65 cycles compared to 25 for Core i5. Combined with the 16KB L1D which means that a huge amount of data will (at best) be hitting the L2 and you've got something terrible going on.
OK, so why are the latencies so high? Here's what I think, you correct me if I'm wrong. Cache latencies are higher when caches are larger, when they have higher associativity, and when they're exclusive rather than inclusive. There is a trade-off between latency and hit rate. The L2 on BD is 2MB instead of 256KB on SB, is 16-way associative instead of 8, and is exclusive instead of inclusive. Hence, the latency is nearly double, and that's very bad.
And I can see why they did it that way. If you want to keep eight threads happy, you have to keep them out of main memory, and the way to do that is with large caches. They didn't want to have a tiny little L2 because there would be too much contention for the L3. They didn't want to have a big L2 which is inclusive with the L3 because it would be a huge waste of transistors to pay for 16MB of cache and only be able to put 8MB in it before you hit main memory. So they went for 8MB of L2 exclusive with the 8MB of L3 -- 16MB total, a good use of transistors, great.
Except that it causes the L2 latency to be terrible, and with a tiny L1 that turns into a catastrophe. I don't know if they didn't notice this or what, but it happened.
Now let's think about some alternatives here. Suppose they adopt the exact same L2 and L3 caches that SB uses: 256KB of 8-way, 11 cycle L2 inclusive with 8MB of 25 cycle L3. That would certainly be faster on anything with a working set that fits in the 256KB L2 regardless of the number of threads, and probably be faster on anything single-threaded because the L3 latency would then be almost as low a the L2 latency is now. And it would save them exactly those transistors that you're claiming disadvantages BD over SB cost-wise. (The disadvantage would be that they would then have a total of ~8MB of cache before they hit main memory, just like SB, and that could impact some highly threaded workloads with large working set sizes. But it seems like a small price.) Adopting this kind of cache architecture for a prospective A-series and selling it into the consumer market along with an integrated GPU sounds like a plan to me, so please explain why it would be unworkable.
Another alternative would be the one I described already: You use the same number of transistors for the cache but you arrange them differently: 512KB as inclusive L2 on each module (or 256MB on each thread), then a 1.5MB exclusive L3 per module, and 8MB of exclusive L4. You get the L2 latency way down for the first 256KB (by making it inclusive and 4- or 8-way associative instead of 16), but you get to keep 14MB of the 16MB of cache as exclusive so that you don't hit main memory on server workloads and you keep anything with a working set smaller than 1.5MB on a single module instead of contending with other m
Nope I dont, but I have sure been in enough research labs. But if you have colon cancer how long are you going to wait to 'educate' yourself? How long do your doctors want you too? What type and how aggressive. Care to be Steve jobs?
And no they don't. They fall for snake oil all the time. ALL THE TIME.
If you have some terminal illness that is killing you so fast that you can't even take two weeks to do your homework and think on it, it seems like the risk:reward for potential snake oil might be quite attractive even thinking rationally. If you're already going to die soon otherwise then what's the worst that can happen?
One of the saddest things I ever heard was about the AIDS precautions taken by haemophiliacs in the late 80's when AIDS was on the rise. People who knew *everything* about blood had the same rate of protected sex as the rest of the population. Nothing like seeing a CDC researcher report that.
I think sex is in a different class from medical treatment. People who have unprotected sex generally don't plan to do it, so it makes perfect sense that people with more information don't make any better decisions. In this case people are making foolish decisions because they don't think, not because they don't know.
OK, so forget the wider internet. Why doesn't the FDA maintain a website containing the approved and known-safe drugs, the experimental drugs, the known-dangerous drugs, etc. The doctor recommends the treatment, the patient goes to the FDA website (which the doctor is required to tell them about) and gets all the information, now the patient can make an informed decision.
You can make all the arguments you want about young children or patients with mental disabilities, but that doesn't justify depriving normal adults of a decision about their own medical treatment.
Slashdot Mirror
Of course. Because Google ads are all over the place, not only on google.com/search.
Should they ever decide to get out of one of those two, I'm pretty sure they'd rather get out of the search business than the ad business.
Let's not be intentionally thick here. Obviously they have more than one web service. But between discontinuing all the services vs. discontinuing all the ads and finding some other method to monetize the services, the latter is obviously preferable.
But the only real monetization strategy for virtually all of those "businesses" is ads. Their business is in building interesting stuff that people want to use, so they can get eyeballs on their ads, so they can pay for building interesting stuff that people want to use, so **STACKOVERFLOW**.
That isn't what's in dispute. But you're confusing the monetization strategy with the business. You don't say that The New York Times is in the advertising business, you say that they're in the newspaper business, notwithstanding that most of their revenues come from ads. Likewise, you don't say that Ford is in the banking industry just because they sell most of their cars on credit.
I love knee-jerk responses just as much as the next guy, but please slowly read again what I wrote.
Does Google make any money from letting you search the web for free? No.
Does Google make shitloads of money from displaying ads next to those free search results? Hell yeah.
So which one is their business?
OK, I'll bite. Let's determine this by taking each one of them away.
If you take away the ads and put up a paywall, do they still make money? Yes. Probably not as much, because paywalls are bunkum, but some significant number of people would subscribe to Google search and they would still make a huge pile of money.
Now, if you take away the search engine and just have a page with ads on it, do they still make money? Of course not. Nobody is going to keep going back to a page that has only ads on it and nobody is going to pay for ads that nobody is looking at.
Most people never imply that when they say that consumers are the "product". Google is not the only company that does this. There are a ton of other companies out there which provide lead generation services and sell existing databases of leads and demographics.
Except that they don't sell databases or demographic information.
In that context, the consumer is the product. Specifically, the information about them.
Except that both of those things are wrong. If the information is the product then you are not the product because you are not information. Moreover, information about users is not either of the products. Advertising is one, web services are the other.
The average user of Google services is the product, the advertisers are the consumer.
That would imply that, having bought you, the "product," the advertiser now owns you. Which is clearly wrong.
That is a contradiction. Which is it? Customers pay money. That's how it works. Otherwise they are not customers. You have buyers and sellers, products, and end users. You can tell which is which by following the money.
That isn't how it works. If I go to a tag sale and buy a gadget, I am the customer. If I negotiate to trade my widget for the gadget, I am still the gadget customer, I am just not paying in cash. It has nothing to do with paying money.
More importantly, your argument is at best that the relation is reversed. However, if the advertiser is the consumer then you must be the seller, not the product, because in either event Google is the intermediary. In addition to that, it still ignores that there are two products (web services and advertising) and that you are still the customer of the web services, even if you are the seller of advertising (which is still silly -- you aren't producing advertising, you're consuming it).
Your mental and semantic gymnastics aside, suppliers don't typically pay retail stores. That is like Walmart charging Chinese companies and manufacturers for the privilege of selling their product. That is not how it works (kickbacks aside). Walmart pays for the product from the suppliers, and then delivers it to the consumer in their retail stores.
Except that as I already explained, advertising is a product with negative value to the consumer. It's like industrial waste. Nobody wants it, so you have to pay people to take it from you.
In order for me to buy into your post I would need to believe that Google is going to ignore the people giving them the money to favor people that do not. Sorry, but that makes no sense.
Your argument flies in the face of all empirical evidence. Advertisers want huge, obnoxious, intrusive ads. Users want small, polite, unobtrusive ads. Advertisers want web pages that keep users there for as long as possible to increase the probability that you'll see or click the ads. Users want to find what they're looking for quickly so they can move on. The users won.
You're so fixated on who pays money that you're losing sight of who has the economic power. The advertisers don't pay a dime if there are no users to consume ads. Which means you have to do what the users want, because losing users means losing money.
Correction: Google's main business is ads. They just happen to go very well with search.
Can we please put a stop to this FUD? You're just repeating Microsoft talking points.
Google does not own you. The only way they get access to your eyeballs is with your consent. You are the customer, not the product. It is 2012, people are not products. Advertising is a product. You consume the product. It is a product with negative value, so Google pays you to consume it, in the form of barter for services. They must provide you with enough value to keep you consuming the product, because you are the customer, not the product. You are the customer of the ads and the customer of the services. The advertisers are not customers, they are suppliers. They supply one of the products, the ads. Google provides the other product, the services. Google is in the web service industry, because that is the product they actually make, and the one that actually has value to you, the customer. There are no ads without services.
It bears keeping in mind why the exchange of services for ads takes place. They could charge money instead. The reason that they don't is that the typical customer values their own money more than they value the cost of consuming that amount of advertising. You would rather consume an ad than pay a penny. The advertiser is willing to pay a penny for you to consume an ad. In consequence, instead of a straight transaction of web services for money, we have a three-party transaction where you consume Google web services and pay for them by consuming some third party's advertising, because that transaction is Pareto efficient over the one where you pay money for services. But that doesn't make the provider of web services an advertising company. They're still providing web services and you are still choosing whether to consume them over the alternatives.
Now, you might point out, Google owns Doubleclick and operates an advertising network. And in that respect their business is advertising -- but that isn't their main business. If Google discontinued all of their web services but continued their ad network as supplied to third parties, they would lose the large majority of their revenues. If they kept the web services and then discontinued their internal advertising network and instead used a third party's, they would keep the large majority of their revenues. What does that tell you about which one is their main business?
How so? The server has to provide the correct certificate before the client sends the User-agent: header. Otherwise, the server would know which site's cert to use from the Host: header, which is sent just before the User-agent: header.
When the user types e.g. 'gmail.com' the default is the http version, which can do a redirect if it detects a browser that supports it. Also, newer browsers can be made to default to https, which solves it neatly because then new browsers default to https and old browsers keep getting http.
Say someone with an SNI browser shares a link with me via e-mail, IM, social network, or however, but my browser doesn't support SNI. Because the person who sent the link is using an SNI browser, the user was redirected to HTTPS and therefore the link is HTTPS. But when I try to open the link, I get a certificate error: "the URL says example.com but the certificate says webhostingcompany.net".
I imagine the same thing as happens if you send them an Office 2007 document when they have Office 2003, or a PDF when they don't have a PDF reader, etc. They type the error into Google, half the results tell them they need updated software, so they download Chrome and it's problem solved.
Such a download won't work for users who aren't the computer's owner or otherwise the administrator.
Chrome defaults to installing in the user's home directory and doesn't require admin rights. I think there is a version of Firefox that does the same thing.
Nor will it work for devices whose browser is part of an operating system package that only the manufacturer and/or carrier can update, such as any Android phone stuck on Android 2.x.
I'm pretty sure you can get other browsers from the market.
Why would people who own a PC based on a P4 or Atom CPU (they're roughly comparable clock for clock) buy a new operating system when they can install Xubuntu?
I don't think they would do either of those things. It certainly doesn't make sense to buy another operating system -- you can get a faster computer with operating system for less than the retail price of Windows.
But as for Xubuntu (or pick your favorite Linux distribution), I imagine it's the same reason people don't do it on newer computers: Lack of awareness, unfamiliarity with a new environment, third party software support, FUD, etc.
And anyway, the point I was trying to make wasn't that SNI would immediately become deployed in April 2014 but that there was a damn good reason not to deploy it before then. Please allow me to correct myself: "This won't change for at least another two years."
I'm still not seeing what requires the wait. The website knows what the client's browser is. There is nothing stopping it from redirecting all users with compatible browsers to the HTTPS version, today. And providing a message to users of older browsers that their shit is old and busted and click here to download the new shiny, without actually preventing them from using the non-HTTPS version in the meantime.
Windows XP isn't going to magically disappear in 2014 when support ends. The number of low-income individuals who continue to make do with some anachronistic Pentium 4 is really quite large, and those people are not about to go out and spend a few hundred bucks they don't have on a new computer or OS just because Microsoft discontinues support. They probably won't even notice, and will replace their OS when the computer dies from being full of dust.
What's going to happen is that XP users will get a message that says "Your version of Internet Explorer is no longer supported, click here to download Firefox or Chrome." And there isn't any particularly good reason for that to happen in 2014 rather than today or in 2016, other than maybe to wait until XP market share drops past some arbitrary threshold.
1) Why would the ISPs care whether the traffic is encrypted or not? It doesn't cost them anything more.
2) All websites are slowly going to 100% HTTPS. The only reason they ever weren't in the first place was that the CPU overhead was too high back when web servers had 1GHz Pentium III Xeons that cost $2000/core. Today practically every chip you can buy has hardware-level encryption support. (There is also an issue about HTTP being a silly protocol that opens more connections and thus requiring more handshakes than it ought to, but that will be fixed soon as well.)
3) Do you recall what DRM is? That's right, it's encryption. What percentage of internet traffic is Netflix again?
Good luck banning encryption when 85+% of all legitimate traffic is encrypted.
Except that...no, not really. Google wouldn't actually know anything more than your ISP (or the NSA) would know by monitoring your traffic, namely, who your friends are but not what you're sharing with them.
There are ways around that. Probably one of the best is to take one of the lightweight boots-in-3-seconds Linux distributions and run OpenGL games in a VM so that you have a native OpenGL implementation. Or even better, do it the other way, run non-Windows as the host OS and then run 'legacy' games in a Windows VM, because the 'older' games will require less fast hardware and therefore care less about any VM overhead.
Another point here is that if you're successful in making OpenGL the standard again, it's very likely that Microsoft will bring back the native implementation on Windows.
More specifically, it's highly likely that you have a crap video driver.
But no PPC support for later browsers will send it to the landfill before that.
--- Eventually we'll be unable to access websites that rely on features in recent versions of flash, java or html5.
You can always put Linux on it. Even the latest Ubuntu runs on PowerPC, which I expect includes an updated Firefox.
The disadvantage is no Flash, but you really shouldn't be running Flash on PowerPC anyway because the latest version has serious security unpatched vulnerabilities. And Flash is slowly disappearing anyway -- your iMac will probably be more useful a couple years from now when Flash is dead than it is now!
A lot of people might like if Steam tried to force publishers to create for an independent platform, but it isn't their job. It isn't even to their benefit, since they seem like they're only creating specifications and not the device itself.
I don't know about that. If they became the new de facto gaming platform, that is a strong position to be in. Especially if the "platform" isn't dependent on any particular vendor (like Microsoft), who may at any time decide that they want to try to eat your lunch.
Anyway, judging from the listed specs, this is not a low-end rig by any measure, meaning that it will already be priced noticeably higher than competitors - so $30 (or whatever it is these days?) for OEM Windows is not likely to make a big difference.
I think this ignores that the goal is to provide consistent specs long-term. The sensible way to do that is to make today's high end the standard, so that it will still have adequate performance going forward, even though five years from now the specs will be met by hardware that sells for a price that makes $35 look significant.
The goal would be to get the entire library of Steam games to run on that console from day 1. It's also easier for the developers if they have, essentially, one platform to target for both PC and console.
My thinking was that you could make Steam the platform, in the same way that Debian doesn't care whether the kernel is Linux or FreeBSD.
The advantage of doing it that way is that you have a sort of "cross-platform platform" -- you say that anything that provides X, Y and Z (e.g. x86-64 quad-core min. 3GHz + GPU of Y minimum spec + Z version of OpenGL + whatever else) is compatible. Then if Apple wants to implement the list of prerequisites on a new Apple TV, or Google wants to implement it on the Google TV, or Canonical wants to make a Steam-compatible media center edition of Ubuntu etc., they can. It lets Valve step in to Microsoft's shoes as the platform maintainer and opens up underlying platform innovation to the world rather than just Microsoft, while still providing minimum standards for developers to target.
Obviously you can't instantly port all of the existing games that way, but there is an easy way around that: You require future games to target the new platform (which Valve can itself implement on Windows, assuming it requires anything not there already), and then put "Windows optional" in the specs, so that if you have Windows then you get all past and future games, and if you don't then you still get all future games and anything anyone wants to port to the new Steam platform in order to get the greater number of sales available on Windows + Everything Else than only on Windows.
My understanding of the story is that essentially this will be a Windows computer with Steam installed.
Is there any inherent reason why it would have to be Windows? If you're going for the console market, i.e. not worried about people needing legacy productivity software, what sense does it make to pay the Microsoft tax? Certainly Sony and Nintendo don't do it, and even Microsoft doesn't run the same operating system on XBOX as they do on PCs.
BSD projects benefit from patches and contributions, both from individuals and corporations.
Exactly, which is why all of MacOS X is BSD licensed. Wait.
You clearly must have missed the part of the GP that says "even being able to take those people don't intentionally contribute."
GPL is perfectly usable in a closed management model when the code is used internally, for example when you provide a service not a software product like google.
Except that if you're concerned about that then you use the AGPL.
Second, it is a political belief, not a fact, that denying access to the close management model is beneficial.
Your opinion regarding the empirical consequences of closed management is rejected on the grounds of it not being a fact.
Your license it for money under an alternative license argument is in conflict with your patches from 3rd parties argument, you can not license code that others own the copyright to - look at the Linux kernel being locked into GPL v2 because all the contributors of patches and new features/functionality can't/won't authorize a switch to GPL v3.
OK, they're in conflict. Which one does that disprove? It's neither of them, isn't it? It's that you get a choice between reincorporating changes from unwilling contributors (without copyright assignment, as with the Linux kernel) or requiring copyright assignment from contributors before incorporating their contributions so that you can license them on different terms while preventing prospective licensees from just taking the code and incorporate it into proprietary software without your consent. You don't get to do either of those with BSD.
Debian packages are not really projects. There are projects which are divided into many smaller packages, like Xorg or Libreoffice, and there are packages which contain many small projects in aggregate, e.g. kdeapps.
That really depends on how you define "projects." It's certainly true that most everything in kdeapps was developed by a common collection of people, but is that really the important dividing line? It seems odd to say that, for example, Konqueror and Kate are not separate 'apps' whereas two pieces of software that provide the same functionality are because they happen to have been written by different people or are more often distributed separately.
Obesity is unrelated to eating disorders.
Well, I think it's pretty obvious that obesity is unrelated to anorexia. The trouble is that you're not just changing the ideal for people who are anorexic, you're changing the ideal for people who are obese as well.
Again, the proper way to deal with anorexia is by educating people on how to live a healthy lifestyle and providing incentives to do so. For that matter, that is pretty well the proper way to deal with obesity too.
Does that help recalibrate your understanding of where I'm coming from? Even a 12x reduction in size, 3x reduction in associativity, a modest "process" advantage, and fewer access ports couldn't give a 2x speedup. You're never going to get even close to halving L2 latency by halving its size. That needs a far bigger hammer.
I understand that. My thinking was that if Intel could manage 11 cycles at 32nm then AMD could too, if no other way than by adopting a cache design similar to the one Intel uses.
Bringing up the idea that Intel's process technology is superior to Global Foundries' rather explains that though, and it gives me another thought: It may be that the reason for the high latencies is intentional in that AMD designed Bulldozer to hit significantly higher clock speeds than it has and high leakage on GF's process is thermally limiting it (and is likewise a significant contributor to the high power consumption).
If that's the case then I think it bodes well for the design, because it means that if GF (or, failing that, TSMC et al) can improve their process technology to reduce leakage, the chips may clock much higher and both performance and performance-per-watt will improve. It also puts an interesting spin on the subject of the article: "Resonant clock mesh" (whatever that is) ostensibly reduces power consumption and heat. If they're successful in implementing it, the design could end up looking much more attractive.
Actually, there is a major benefit to a processor which is 50% faster, even if it's idle much of the time. Power efficiency. There is increasing focus on total energy consumption, to improve the battery life of portable computers. CPU usage profiles for light loads look like "wake up, do something, go back to sleep". If you have two CPUs which both use the same power, but one is 2x as fast, the faster CPU's time spent in wake is literally half as much as the slower CPU. Now throw power gating into the mix -- when idle, modern CPUs can literally be shut down almost completely, consuming no power. This means the 2x faster CPU is literally 2x as energy efficient, even for light loads, because they both consume the same amount of power while active, but the faster CPU is active for much less time.
That is true, though I'm not sure how large a loss that would be in practice: It has minimal effect on the desktop market and portable computers of the "slower CPU / faster GPU / low watts" variety are the wheelhouse of the Bobcat rather than Bulldozer, which takes care of anyone whose primary concern is battery life. Then AMD can still get anyone who prioritizes purchase price over battery life, as well as anyone who prioritizes battery life and better-than-Bobcat performance over weight (since you can use the money you save buying AMD over Intel to install a bigger battery that more than makes up for the power consumption). All it really excludes them from is those who prioritize performance, battery life and weight over price, which has a pretty high overlap with the market that demands the highest single-thread performance at any price (which they had already lost).
Wait until we have real universal health care.. then the anorexics health becomes everyones business... and thus, justifying laws like the banning of skinny models. Think of this research like a wake-up call.
The biggest problem with this isn't even the "big government" problem. It's that the researchers are empirically wrong about the solution.
Are we not presently in the middle of an obesity epidemic? The problem is not that people want to be thin -- that would probably be a good thing. The problem is that they're doing it wrong, i.e. in an unhealthy way rather than through proper diet and exercise. So how can the solution be to make people not want to be thin anymore? That would just result in more couch potatoes and more heart disease.
The solution is to see to it that the people who are dedicated to looking fit are educated about how to do that, i.e. a combination of sit-ups and peanut butter rather than nervous tension and starvation.
Yeah, that's pretty much it. There is no "cyber war" -- but there is state-backed industrial espionage.
And the problem with the way these people are thinking about it is that MAD is totally wrong. It isn't offense that you need, it's defense. Especially for the US: What "the enemy" is doing is sabotage and misappropriation trade secrets on a massive scale. Even if you can manage to do the same thing to them, your infrastructure and trade secrets are more valuable than theirs. Which means that having the same offensive capability as they do means that you lose.
You can't win unless you can prevent yourself from losing. The way to prevent "cyber war" is to keep sensitive industrial equipment disconnected from the internet, provide incentives for relevant people in industry to implement security best practices, etc.
The idea of creating an arsenal of "cyber weapons" is fairly preposterous. About the only good value I can imagine in having one is that it gives you a good threat assessment for things you need to be able to prevent when designing your defenses. The idea that there is ever going to be some kind of doomsday cyberwar where they launch a "cyber attack" to shut down the power grid or some such Hollywood bullshit is totally inane -- even assuming that the enemy launches such an attack, how are your "cyber weapons" going to do anything in response when your country no longer has a working internet? In other words, you need defense, not offense. And if you have a sufficient defense, the offensive capability is superfluous.
I'm going to go ahead and concede that I don't design microprocessors for a living, and I'll give you the heritage of the various architectures because I'm primarily just reiterating what others have told me and I feel like at this point unless you have some kind of non-public information we're both just speculating about it anyway.
So if you want to be the expert, let's see if we can learn anything here instead of just having a flame war.
You're arguing that Bulldozer uses too many transistors for its performance, and never mind that most of those transistors are wasted on an unfortunate cache architecture because it would be too difficult to change that. At the same time, you seem to think nothing of Intel basically starting over entirely with the L2 between Katmai and Coppermine. So let's see if you can help me understand why that is.
If you look at the cache latencies for Bulldozer, they're terrible. The L2 is 21 cycles compared to 11 for Core i5. The L3 is 65 cycles compared to 25 for Core i5. Combined with the 16KB L1D which means that a huge amount of data will (at best) be hitting the L2 and you've got something terrible going on.
OK, so why are the latencies so high? Here's what I think, you correct me if I'm wrong. Cache latencies are higher when caches are larger, when they have higher associativity, and when they're exclusive rather than inclusive. There is a trade-off between latency and hit rate. The L2 on BD is 2MB instead of 256KB on SB, is 16-way associative instead of 8, and is exclusive instead of inclusive. Hence, the latency is nearly double, and that's very bad.
And I can see why they did it that way. If you want to keep eight threads happy, you have to keep them out of main memory, and the way to do that is with large caches. They didn't want to have a tiny little L2 because there would be too much contention for the L3. They didn't want to have a big L2 which is inclusive with the L3 because it would be a huge waste of transistors to pay for 16MB of cache and only be able to put 8MB in it before you hit main memory. So they went for 8MB of L2 exclusive with the 8MB of L3 -- 16MB total, a good use of transistors, great.
Except that it causes the L2 latency to be terrible, and with a tiny L1 that turns into a catastrophe. I don't know if they didn't notice this or what, but it happened.
Now let's think about some alternatives here. Suppose they adopt the exact same L2 and L3 caches that SB uses: 256KB of 8-way, 11 cycle L2 inclusive with 8MB of 25 cycle L3. That would certainly be faster on anything with a working set that fits in the 256KB L2 regardless of the number of threads, and probably be faster on anything single-threaded because the L3 latency would then be almost as low a the L2 latency is now. And it would save them exactly those transistors that you're claiming disadvantages BD over SB cost-wise. (The disadvantage would be that they would then have a total of ~8MB of cache before they hit main memory, just like SB, and that could impact some highly threaded workloads with large working set sizes. But it seems like a small price.) Adopting this kind of cache architecture for a prospective A-series and selling it into the consumer market along with an integrated GPU sounds like a plan to me, so please explain why it would be unworkable.
Another alternative would be the one I described already: You use the same number of transistors for the cache but you arrange them differently: 512KB as inclusive L2 on each module (or 256MB on each thread), then a 1.5MB exclusive L3 per module, and 8MB of exclusive L4. You get the L2 latency way down for the first 256KB (by making it inclusive and 4- or 8-way associative instead of 16), but you get to keep 14MB of the 16MB of cache as exclusive so that you don't hit main memory on server workloads and you keep anything with a working set smaller than 1.5MB on a single module instead of contending with other m
Nope I dont, but I have sure been in enough research labs. But if you have colon cancer how long are you going to wait to 'educate' yourself? How long do your doctors want you too? What type and how aggressive. Care to be Steve jobs?
And no they don't. They fall for snake oil all the time. ALL THE TIME.
If you have some terminal illness that is killing you so fast that you can't even take two weeks to do your homework and think on it, it seems like the risk:reward for potential snake oil might be quite attractive even thinking rationally. If you're already going to die soon otherwise then what's the worst that can happen?
One of the saddest things I ever heard was about the AIDS precautions taken by haemophiliacs in the late 80's when AIDS was on the rise. People who knew *everything* about blood had the same rate of protected sex as the rest of the population. Nothing like seeing a CDC researcher report that.
I think sex is in a different class from medical treatment. People who have unprotected sex generally don't plan to do it, so it makes perfect sense that people with more information don't make any better decisions. In this case people are making foolish decisions because they don't think, not because they don't know.
OK, so forget the wider internet. Why doesn't the FDA maintain a website containing the approved and known-safe drugs, the experimental drugs, the known-dangerous drugs, etc. The doctor recommends the treatment, the patient goes to the FDA website (which the doctor is required to tell them about) and gets all the information, now the patient can make an informed decision.
You can make all the arguments you want about young children or patients with mental disabilities, but that doesn't justify depriving normal adults of a decision about their own medical treatment.
Do you have cancer? Because I bet cancer patients keep up with it.