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


Forgot your password?
Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. ×

Comment Re: Acceptable Ads (Score 1) 523

personal home pages

That was for the most part geocities, and there was a wealth of technical information there. Oh, and it was ad supported.

Sure, I've never claimed that people don't use free stuff if they can get it and I'm also not saying that people should not display ads if they like them. I just happen to dislike them.

But I've also never heard of any ISP then and now who didn't offer some space for your web pages. In my opinion, Geocities was not popular primarily because they offered web space but because the neighborhood metaphor that gave them some special flair (plus, something like an integrated webring). It was fun browsing around and putting stuff on it that you wouldn't want to put on your primary homepage. But it wasn't more than that. I was there around the same time.

And you can hardly deny that most of the useful web sites are only useful because they provide user-generated content that formerly could be found on usenet and various university or personal homepages.

Last, hotmail... sorry, but that was always crap. Just like AOL.

Comment Re:Ads are not acceptable. (Score 5, Interesting) 523

You're the clueless idiot. I can display web pages in my browser as I like. It's my machine and I pay for the bandwidth. I have zero obligation, neither morally nor legally, to watch advertisements or even display them, just as I don't have any obligation to click on links in spam mails sent to me.

Moreover, I don't have to discover new products. When I want to buy something, I inform myself and then buy the product that best fits my needs. And I am seriously not interested in the flawed business models and whining of self-proclaimed entrepreneurs who have no genuine product to offer.

Comment Re:Ads are not acceptable. (Score 2, Interesting) 523

Total bullshit. I've said it elsewhere as AC, but let me repeat it:

The fallacy of many web-sponsored startups is to believe that their "content" is good or even worth anything, just because people look at it for free. Mostly it's not. (There are exceptions, of course.) If Facebook would die tomorrow, nobody would give a shit about it, people would simply move on to another site. The same holds for most of the other adware sites. If you have a good product, people will buy it. Ad-supported "content" is just a soap bubble.

Comment Re: Acceptable Ads (Score 3, Informative) 523

There was no problem getting such technical information before the Web was commercialized. In fact, Usenet, FAQs, personal home pages and swapping a few emails with other enthusiasts are usually better for this purpose today than any ad-driven sites you can come up with. Most technical information sites base their "content" on the free work of their users anyway, which is exactly what Usenet was and is intended for, plus its decentralized by design.

Comment Re:Privacy Badger and uBlock Origin? (Score 1) 523

I'm using uBlock Origin and think it's better than ABP, because you can easily pick elements of any page for cosmetic blocking. For example, I've created my own rules on the fly to block sneaky ads at Arstechnica and unwanted youtube "recommendations".

However, it seems to me that its easier to detect by anti-adblock services, but perhaps that's also just because the advertisers have improved their techniques. Hard to tell, as I haven't used ABP in a while.

I block absolutely everything that even remotely resembles an ad, and don't want anyone to earn single dime from my website visits.

Comment Re:No, here is what they need to do first (Score 1) 748

Haskell has been used for safety critical code in the aviation industry

Do you have a link I could read about it? My google-fu is weak.

I honestly thought that I'd heard that at least one big player in the aviation industry used Haskell, but after looking around myself, I believe you're totally right in asking for some evidence. Perhaps I've fallen prey to some Haskell aficionados who wanted it to be used for safety-critical systems. But e.g. these guys:

only use its type system for their own language, and only for research purposes. So yeah, sorry about that.

I never claimed that.

Then go back and re-read what you wrote. It's ambiguous and my interpretation is completely fair.

Yes, I also want to apologize for that, although you could perhaps have given me the benefit of the doubt. This list was meant as an example of technologies that are dubious from the perspective of safety-critical systems. Machine learning definitely belongs into this category. BTW, just to rule out another potential misunderstanding, in the context in which I am using the word "deterministic", it means fixed runtime and memory consumption guarantees within a given system (including the hardware).

You seem deeply confused.

Yet perhaps I only seem so. Anyway, better not write things like that! It will antagonize people needlessly.

You're going on about how machine learning shouldn't be used. Given that before machine learning was up to the task, no reasonable self driving cars existed AND given that all existing self driving cars use machnine learning, the onus is on YOU to provide a convincing argument that machine learning is unnecessary for making a self driving car.

Why do you think so? My point is that any technology that potentially endangers the lives of other people should satisfy some stringent safety requirements. Nobody would argue in the aviation industry "look we can only make it fly by using inherently unsafe and unpredictable technology, so that's what we should do". On the other hand, I've worked in the aviation industry, so perhaps I am being too optimistic...

Formal code verification, at least for safety critical parts of the code, is commonplace in the aviation industry

I don't think you know what formal verification actually is and what it can do.

Fine. Your opinion. Luckily that doesn't affect my knowledge in any way.

You can prove the algorithms terminate and prove they don't leak memory and prove a bunch of things.

Exactly, and that's what I want them to do, before unleashing their cars on millions of pedestrians. That's a very simple and coherent position.

What you can't do with formal verification is prove it actually detects pedestrians. Which is kinda critical.


It might come as a surprise to you...

*chuckles* No, that didn't come as a surprise. All I want from the makers of self-driving cars is that they use some safety standards that have been accepted in the aviation industry despite being expensive. Frankly speaking, it kind of weird to argue against this, given the fact, which you have pointed out so eloquently, that autonomously driving a car under varying conditions is a way harder task than autonomously flying a plane. But I do understand that monetary concerns would speak against higher safety standards ...

Comment Re:No, here is what they need to do first (Score 1) 748

Write all the code in Spark or at least Ada or another language with a similar safety record (e.g. Haskell, perhaps Rust).

Haskell? Are you nuts? Much of the code needs to be hard realtime. That rules out garbage collection completely.

Haskell has been used for safety critical code in the aviation industry because its type system makes automatic verification easier. Whether hard realtime is needed depends on the specification of the respective component. If it needs to be hard realtime, you cannot use GC or dynamic memory allocation. That's plain obvious.

2. Use only deterministic code, no dynamic memory allocation, fluffy A.I. heuristics or machine learning. I don't care how hard it is.

1. Machine learning is not the same as nondeterministic.

I never claimed that.

2. Machine learning doesn't mean non guaranteed response times.

Never claimed that either. I don't what makes you think that I believe that machine learning should be used. It probably shouldn't.

3. What you are suggesting isn't just hard, it's impossible.

No of offence intended, but this claim makes me doubt whether you know what you're talking about. Formal code verification, at least for safety critical parts of the code, is commonplace in the aviation industry. You probably thought about the Halting Problem and believe that it somehow shows that no program can prove that another program is correct and will halt. Well, it doesn't. Formal correctness proofs are done all the time for Ada/Spark programs, that's what the additional annotations are for.

You know, people have been working on self driving cars for years.

Nothing new here either.

My point of view is that the safety standards for self-driving cars are not yet at the level at which they ought to be before the technology is released to the rest of the world. If you disagree, fine - not my problem.

Comment No, here is what they need to do first (Score 1) 748

1. Write all the code in Spark or at least Ada or another language with a similar safety record (e.g. Haskell, perhaps Rust). It can be a formally sound and static subset of C++, if it must be, but not just any C or C++.

2. Use only deterministic code, no dynamic memory allocation, fluffy A.I. heuristics or machine learning. I don't care how hard it is. I want real engineering based on real physics with provable security margins, guaranteed response times.

3. Formally validate the code in a theorem prover, as it is done for planes.

4. Let the whole design be reviewed and checked by external expert commissions, according to previously defined safety standards.

After that, we can discuss whether they are allowed to break the rules.

Why all that? Well, nobody should give a damn about safety statistics compiled by the companies who produce the self-driving cars, and it doesn't even matter whether they are safer than human drivers or not. What matters is that these companies have tons of lawyers, whereas the end-consumer has not. So good luck if you have an accident and it could be their fault. There needs to be some tight regulation like in the aviation business.

Slashdot Top Deals

He who has but four and spends five has no need for a wallet.