Catch up on stories from the past week (and beyond) at the Slashdot story archive


Forgot your password?

Comment The (linked) Aandtech article on battery life... (Score 4, Insightful) 558

The (linked) Aandtech article on battery life pretty much answers its own question.

Surface pro and surface pro 2 completely destroy everything else in the benchmark ratings. It means haswell doesn't manage lower power scenarios nearly as well as ARM, but Intel never has.

For a comparison to iOS they'd need to well, actually have on on their chart. I can certainly see the argument that Windows is worse at power management than other OS's on the same hardware - but without hard numbers in a chart that's a tough case to make, since you're comparing different review sites to each other. Comparing different hardware is missing out on a lot - for most computing needs they're benchmarking Haswell is massive overkill - which might just be it, it literally cannot slow itself down enough (with either MS or intel drivers being the culprit) to save even more power.

Or windows is doing background stuff that other OS's aren't. Whether those provide any value to justify reduced battery life or not is debatable, but the answer seems to be 'probably not'.

It still isn't 'microsofts hardware', it's hardware from some 3rd party vendor they soldered together in a case and put their own sticker on it. Yes, it's up to MS to try and ride the cases of Intel and whomever is supplying their displays and SSD's to find ways to save power, but it's ultimately up to the 3rd party guys (who also sell parts to the rest of us) to actually make the drivers for their hardware.

Comment Re:routine IT work (Score 1) 307

... it's called fuzzing.

And Fuzzing sucks, because you're making assumptions about how much to add still. And your assumptions can easily turn out to be wrong.

Real data isn't necessarily random - that was the example I gave you for a reason. Systematic biases introduced into real use you can't simulate for. Random usage patterns you can *try* and fuzz, but you can easily screw that up, depends on the system.

again, this isn't new.

No it definitely isn't. This sort of stuff has been around hundreds of years, and if you think 'fuzzing' will deal with systematic biases in your data you should probably come back to school, we'll sort you out when you get to second year.

Computers are fast.

But finite. That you can do with 10 computers what took 100 before doesn't change the problems caused by having half (or 1/4) the number of computers that you actually needed.

The moment you cannot fit the entire problem on one box with room to spare you are into all of the problems of running out of capacity, which are not new.

Comment Re:routine IT work (Score 3, Insightful) 307

We keep saying it over and over because:

It's easy.

Isn't true, at all. Watch 1000 volunteers all try and test your system, and then try and simulate that behaviour to model however many you think you'll have - and you still get surprised.

Yes, definitely you should have testing for all the cases of what a user can do. But you don't know how people are really going to use a system until they're using it for real. As it turns out real use is different than testing, and an early tester sample are not really a good sample.

Let me give you an example of how synthetic tests will go badly. You get some fake numbers from your partners so they all have either completely random prices, or they all have exactly the same price. So when you're testing users click around, and no problem right, they can select the one they want etc.

Then you get real data, in the real world - and one company posts a price 3% lower than the other guy but it's for a slightly different product. So before, where users clicked the same thing once, now a bunch of them are clicking back and forth loading the page multiple times doing so, they're sending them to their friends to compare etc.

It then takes your assumption about 100 000 users per box to maybe 70, 80 or 90 or some number not quite enough when you scale up to 5 million.

Comment Re:On the other hand (Score 1) 178

Ivy bridge was a bad best option. Haswell is basically the same performance for 40% less power consumption but didn't exist back when they made surface pro 1.

Microsoft isn't a chip maker - there was literally no good option for a CPU for the surface pro that would be x86 - AMD processors were all as hot or had bad performance.

ARM, sure, they had choices, but go build yourself a PC and try and find an x86 cpu that isn't from Intel or AMD and is reasonably priced and works worth a damn.

Comment Re:routine IT work (Score 1) 307

While that's certainly true, designing a system for 5 million users versus 10 million users is not radically different - but if 10 million users all try and hit your server designed for 5 million you're going to have problems.

There's no way was done as a 'hobby' type site - they guessed wrong on the load they were going to have and there seem to be some issues with how the insurance companies hook up to the system - neither of which is encouraging but it's not like they intended 100 users and got a million.

Although it would be interesting to know if some of this is largely a 'first day' sort of problem, where the first few weeks just has way too many users and by mid november it will be more or less settled. And then all the people who didn't sign up will panic and need to over christmas and there will be problems again.

Comment Re:I've got a better idea (Score 4, Insightful) 307

And there's no data like real world data for load. At some point you sit down in a room and guess how people are going to use your software. You put it out there, and find out you were wrong.

That is after all, why they did this with a couple of months to spare.

Are there going to be 10 million people over christmas all trying to buy health insurance? Probably, and that's going to cause no end of grief, but there isn't some mystical open source fairy that can tell you how to correctly predict load for a system like this and make all the infrastructure work the way you want it to. Particularly with health care and open source you'd have to deal with thousands of tea party programmers trying to fuck it up too.

Comment Re:I can't tell from the post (Score 3, Insightful) 105

Those wouldn't be starting salaries unless you have a graduate degree or experience and are jumping into a senior position.

Last I checked (which was a couple of years ago) a fresh PhD starting at google was in the 130-140 range, and a fresh undergrad was usually in the 80 range - at least that's what my students get.

Comment Re:Bad data (Score 3, Interesting) 105

Don't rent buy.

That's same with any big city/small town setup - you buy a house in the city (or suburbs of the city anyway), when it comes time to retire you sell your house and move to a small town half an hour a way, buy a comparable house for half the price, and the other half becomes most of your retirement nestegg.

Comment Re:Deep down.. (Score 1) 610

Yes, because we have Bush light as Prime minister. But he's wrong about trying to climb further in bed with 'murica. And that's my point - just because he's wrong about it doesn't mean he's going to change. The people who have power and sensible policies are as much going along with NSA spying as people with bad policy. So the options are to find fringe crazy people who are only right about NSA spying and wrong on everything else. And on the scale of things NSA spying is a far less serious problem than a lot of other things.

Comment Re:Deep down.. (Score 1) 610

Well the perps were allegedly outsiders and the NSA's job was to spy on outsiders.

So was the CIA, and they dropped the ball too.

No intelligence agencies had much success at trying to infiltrate Al Qaeda, and most of them were only making a half assed effort to do so at best anyway. Hence the need for more resources.

That doesn't mean a massive data centre in utah is the right resource - but in addition to spying on China and Russia and India and China and France and the UK etc. the US now needs to contend with another player in the field.

Comment Re:Deep down.. (Score 1) 610

So they are violating the law and the constitution, and just because we've thought it's been going on for a long time, now that we have proof we should just let it go?

So, aside from the obvious, that constitutions are stupid and they are chartered by law to do what they are doing - my point wasn't that what they're doing is good. It's that it isn't surprising.

It's hard to work up outrage over a program that has been going on for over 20 years, that everyone with any influence has known is going on for 20 years. That doesn't mean being in favour of it.

Slavery was not legal in the UK from 1772 on - but it was legal in the colonies until 1833 (although the slave trade was abolished in 1807). That's a long time for something obviously abhorrent to be going on. But you can't be out protesting every day for 60 years about something you disagree with. Until someone comes along that looks like they're actually going to do something about it you aren't going to change the law and keep your job and your family fed at the same time. I'm not against ideological crusaders jumping up and down about things - that's how you get change. But if the choice of whom to support is Ron Paul (or his idiot boy son) then the US is better off being spied on than that lunatic being in charge of anything.

The people in charge in basically all of the NATO countries, Australia and New Zealand are all in favour of the program, so... I might not be in favour of it, but I'm also far more worried about whether or not government policy is going to actually do something about the economy, healthcare etc. So given the choice, and the lack of political will to do anything about it, I'm not going to be out protesting all the time. My country, Canada, should already be walling itself off from the US (both literally and figuratively), NSA spying is only a sideshow to that as a broader political problem. If you're in the US, healthcare is far more important than NSA spying. You can't worry about being spied on if you're dead, and if you're bankrupt due to getting cancer you have bigger problems than whether or not the government can collect you e-mail contact list.

Comment Re:Deep down.. (Score 3, Interesting) 610

Every one of the 9/11 terrorists fit a profile that should have sounded alarm bells at the border.

I hate to break it to you, but half the male muslims in the world meet the same profile. And the vast majority of them don't try and crash airplanes into buildings.

But that's beside the point - I didn't any of it was a good idea. Only that it's not surprising.

Russian operatives were far more successful, some escaping detection for multiple decades.

Yes, but spies are professionals.

And enough Al Qaeda operatives have escaped detection to cause quite a lot of trouble (including incidentally in Russia, which unlike the US, has internal border controls as well).

Slashdot Top Deals

Everybody needs a little love sometime; stop hacking and fall in love!