Become a fan of Slashdot on Facebook


Forgot your password?

Comment Re:We put everything in AWS (Score 3, Informative) 237

It can be for cost savings if you have highly bursty workloads. If you need N machines for your peak loads but 0.1N machines most of the time, then something like AWS and only paying for 0.1N machines except during demand spikes is a lot cheaper than hosting N machines yourself. If you have a consistent workload, hosting your own machines is probably noticeably cheaper.

Comment Re:Psst... Don't tell anyone (Score 3, Insightful) 237

This is exactly the sort of negativity which shows how the open source community is abusive and unable to cope when a great new idea comes along that throws away all those bad concepts in Unix, just because we're right and you're wrong.

Nonsense. Compare the reaction to systemd with the reaction to launchd (XNU) or SMF (Solaris). Most people who have had contact with either of the latter regard them as imperfect but significant improvements on what was there previously and, if they're using systems that don't ship with them wish that they did (or, ideally, something taking the good ideas from them each and combining them, leaving the bad ideas behind). No one is complaining about replacing traditional UNIX tools with something better, they're complaining about replacing stuff that mostly works with something that throws all of the last few decades of software engineering away.

Comment Re: no extra calories? (Score 3, Interesting) 363

I think there may be other factors.

I'm not dieting.

I exercise modestly (a mile or two a day walking and 30 pushups and giving a couple hours of therapeutic body work a week).

My weight and blood pressure have dropped since I retired (at 51).

Blood pressure from 160ish to 112 last visit. Weight from 278 to 245.

I have 2-6 artificial sweeteners a day in my coffee/soda and I also cook with it.

My blood sugar has declined from 144 when I retired to 112 last doctors' visit a couple months ago.

I occasionally still go to buffets.

Perhaps stress contributes. Work was killing me and I'm not the only person I know like that. One friend dropped 5 of 6 blood pressure medicines in the 6 months after he retired.

Comment Re:The libraries we choose (Score 4, Insightful) 237

Uh, isn't that true of any native library? Most C and C++ libraries don't do anything you couldn't do yourself

You've selectively quoted him. The full complaint was:

many people use jQuery (which is a large CPU-heavy library) to do things that can be done in fewer lines of straight javascript

Why use a library and 10 lines of library calls to do something that you could do in 5 lines of code? You should use libraries when the cost of reimplementing the functionality is higher than the cost of using the library.

Comment Re:More on this please! (Score 1) 488

There's a lot of ongoing work. Some of it is in formal models of programming languages: we had a paper in PLDI last year on a model for C (which has to be parameterised, because it turns out the C standards committee, C programmers, and C compiler writers have almost totally disjoint ideas about what C means). We're looking at combining this with formal models of CPUs to allow black-box testing of compilers (if we can prove that the output corresponds to the input then we don't need to verify a few million lines of compiler). There's white-box stuff, such as the Alive project from Utah and MSR looking at proving correctness of a number of the peephole optimisations in LLVM (although, unfortunately, it's so far managed to prove incorrectness of a depressing number of them).

Beyond that, there's a lot of ongoing work improving the proof tools. Coq, Isabelle and HOL4 are still in active development and AGDA has gained an increasingly large following in recent years. The more that you can automate the proofs, the more feasible this kind of thing becomes and, in particular, the easier you can make it to write modular proofs the easier it is to maintain the code. The CEO of Intel just after the Pentium FDIV bug started to pour money into verification and made quite an insightful comment in this regard: He didn't care how much it cost to verify something the first time (it had to be cheaper than doing a full recall), but he needed it to be cheap to re-verify after some changes. Centaur (formerly VIA's x86 chip division) incorporates a lot of verification into their CI system: they're a long way away from full verification, but their verification tools help catch and pinpoint a lot of things that would otherwise (probably) show up in testing in the simulator as 'something is broken in the design'.

There's also a growing push towards proof-carrying code. The F* language (MSR and INRIA) is probably the best example in this space. It's a version of F# that allows you to write proofs in the same language that you write the code and only compiles when the code matches the proofs. Unfortunately, F* still has a very long way to go before simple changes are easy. For example, they have a talk where they walk through verifying a quicksort implementation. If you want to change the quicksort to a heapsort, then you have to throw away most of the proof (and the proof is more complex than the implementation, though at least the input / output specifications are simple and so you can be sure that the proof is correct if it compiles). They're also using an SMT solver (Z3) on the back end, so there's no guarantee that it will actually terminate: you will either get a counterexample, a success message, or a timeout in the solver and a timeout doesn't necessarily mean that the proof is correct or incorrect, just that it's too complex to machine check.

Comment Re:Does formal verification really prove correctne (Score 1) 488

Formal verification proves that the program corresponds to the specification. It doesn't prove that the specification is correct. It's really hard not to run into Goedel's incompleteness theorem and discover that the specification is actually more complex (and therefore error prone) than the program.

Comment Re:Double Checking (Score 1) 504

Your number is really high. Solar cells are much more durable. And since they are not moving, I find the bird strikes, frankly, to be quite hard to believe.

And the new panels are hail resistant.

A more realistic figure would be 20 years and even that is being a little pessimistic.

AND, when solar panels hit the right price point- growth is going to be phenomenal. Germany with about a fifth of our population has powered the entire country on some days after only 4-5 years of building.

And many of those solar panels will be local installations scattered all over the country.

And some of them may be molten salt plants instead of photovoltaic.

Comment Re:Too bad. (Score 4, Insightful) 397

At the same time, odds are at least some are being hit up for more than they owe, perhaps being double billed. There may even be some that actually don't owe any money. We can't tell how many because the paperwork is too screwed up to show that the loans even exist.

Since these loans were bought, it is safe to say that none of the debtors agreed to owe the current holder of the loan.

It's more interesting to pull back a level. There we see a financial industry so high on itself that it figured it didn't NEED proof of anything. Just point and say "He owes me money" and the courts would oblige. Sadly, they weren't wrong at the time. But after screwing around for 10 years soiling their own reputation at every turn, it is no longer true. Now judges want to see proof (like they should have all along).

Comment Re:An error? (Score 1) 142

I can think of two ways. Way one: you contract out the development to a company that has a paid version and an ad-supported version. They accidentally give you the ad-supported apk to push to your customers instead of the paid one. Way two: you have an internal project to see if you can push down the up-front cost of a phone by pushing ads in various places (as the Kindle Fire does on the lock screen) and using that to subsidise the cost. You have the same people working on it as the normal system and one of them accidentally pushes to the wrong repo.

Slashdot Top Deals

If the aborigine drafted an IQ test, all of Western civilization would presumably flunk it. -- Stanley Garn