What's really at issue, here, is that the firm wants to get the benefit of the safety testing, but not only doesn't want to pay for it, but also wants the FDA to act as a sort of a legal shield in case somehow they screw and up still wind up selling a cow with mad-cow, even though the beef was 100% testing.
You're obviously speculating wildly without doing your homework. Your claims are at best unsupported, and your claim that they don't want to pay for the testing is directly contradicted in the blog entry and legal brief in the following respective paragraphs:
A Kansas-based exporter, Creekstone Farms Premium Beef, seeking to test its cattle to minimize public fear, challenged Department of Agriculture regulations that block corporations from buying and using kits to test for mad cow disease. There is no cure and no treatment for the neurological disease. Itâ(TM)s 100 percent fatal.
The United States
Department of Agriculture (USDA), however, asserting
authority under the Virus-Serum-Toxin Act, 21 U.S.C. ÂÂ 151-
59 (VSTA or Act), denied Creekstoneâ(TM)s request to purchase or
use a BSE test kit.
So now Granny calls for support. She can't uninstall. She is going to have to do a manual uninstall. So we email her a document with a procedure to run the computer in safe mode. Delete a bunch of files and folders AND then run regedit and pound a bunch of entries out of the registry
Trust me. I took at least 2 or 3 calls like that a day.
You filed at least two P1 bugs right?
Installer broken when resources insufficient. (It should complete at the lowest RAM/disk/whatever you've put through QA and print a clear error message otherwise, leaving the system in its original state.)
Uninstaller broken when installation half-finished (or in safe mode, or whatever it was that made you shift into "manual uninstall").
and mentioned that they were definitely costing the company $X,XXX/day in support and probably costing more in lost customers as bad word-of-mouth spread?
Arguably the others - broken installer system (Microsoft's bug), bad hardware, and pre-install virus (which may not have existed when the media shipped) are not defects in your product, but they're costing you money as well, so they should be worked around if at all possible.
It'd be easy to write a small program that requires little RAM and no disk (beyond the binary) which would automate at least the "delete the files and do this in regedit" portion of your instructions to grandma. Ship it with the media and (for customers with Internet connections and installations predating the fix) put it on your website. Call it "uninstaller version N+1".
Working around serious software defects should not be considered an "ordinary" task.
With all due respect but disregarding what good prison would do this kid as "complete ignorance" is ignorance in itself. The kid is 18, even if he saw just a year of jail time, you mix him up with all kinds of career criminals you end up getting out something worse than what you put in.
I hope you're saying that because he's somewhat young and the crimes were (presumably) his first and non-violent, not because most were computer crimes (a "clever" white-collar crime many here could identify with). Why not apply the same reasoning to, say, a high school student who commits a string of burglaries/thefts?
I wish to download your proxy server, and subscribe to your newsletter.
All right. It's a work in progress (no keepalives yet; requires libevent trunk + a few not-yet-merged patches), but I'm willing to share what I have. I've traveling now, but I'll put something on my website Monday night. I'll call it FoxIgniter, unless someone has a better name.
Argh! Yeah, I hate that bug ("[PAC] Defer proxy resolution for HTTP and HTTPS PAC to avoid blocking main thread during DNS resolution")!
I wrote a proxy server to run on localhost, do DNS resolution, and send the request to the appropriate upstream proxy or directly to the source, just to work around this bug.
Do you realize how hated Hillary Clinton is among republicans? Quite possibly the most polarizing figure in politics right now. They see her as disingenuous and power-hungry, and quite frankly she is. There is absolutely no way, by any stretch of the imagination that Hillary Clinton is more electable.
Many Democrats would now agree with you and the Republicans that Hillary Clinton is disingenuous and power-hungry.
Certainly she was disingenuous in her claims of winning the popular vote. She included the votes in Michigan, where Obama followed rules by withdrawing his name but she did not. She supposedly did this to because every vote should count, but her statements showed no regard for that before she fell far behind. At the same time, she's certainly not counting every vote by arbitrarily excluding all caucus states from consideration.
Power-hungry in that her actions have led us to fear she'll tear the party apart and give ammunition to the Republicans rather than admit she's lost. Again with Michigan. Her lawyers' statements make me fear she'll launch a legal challenge to the assignment of Michigan's "uncommitted" delegates to Obama. As irregular as it was, the only better way to represent the wishes of those voters would be to have a new vote. I imagine she'd oppose that because Obama would win it. (Even ignoring consideration of the events between January 15th and now, I think it's fair to say that if Obama's name were on the ballot, more people would have shown up to vote for him than actually did show up to vote for "uncommitted". Given that 40% of the people who did show up voted for "uncommitted" or "undecided", that likely means he would have won.)
I'd also throw in "pandering", particularly in her suggestions of foreclosure and gas tax holidays. ("I'm not going to put my lot in with economists.") The only other explanation I see (that's she's not only missing the obvious flaws in these schemes but also rejects the concept of people who devote their professional lives to studying the issues having a better understanding of them) is not a good quality either.
I had a sliver of doubt when I cast my vote for Obama instead of Clinton in California's Democratic primary. That sliver is long since gone.
Leaving race out of the issue, how many republicans do you think would vote for someone named Barack Hussein Obama. A name that rings with the sounds of two recent so called enemies.
We should vote for someone else because other people will vote for someone else because his name reminds them of two bad people?
I want to live in a world where people don't cast their votes for such stupid reasons. And to quote Mahatma Gandhi, "you must be the change you want to see in the world." If that makes me too idealistic, so be it. Better that than too cynical like you.
If the FCC strikes down cancelation fees then the price of phones will suddenly increase several hundred bucks. This isn't necessarily a good thing for the market since almost everyone I know tends to go for the free phone or the 50 dollar phone when getting a new plan - no one is willing to spend several hundred dollars. At least, not in a lump sum up front.
Right; I'm not willing to spend several hundred dollars in a lump sum up front. That's because I'm paying hidden monthly fees for the phone already. If I break the contract, I'll pay an early termination fee and be left with a useless service-locked phone, and my next contract will inevitably pay for a new phone anyway. So what would I gain by paying money up front, too?
Striking down cancellation fees and allowing open access (any phone on any network) would change the game. I'd gladly pay for my phone outright.
But let's assume you're right - even without those gotchas, most people would want the most expensive phone on an installment plan. If the carrier early termination fees go away, cell phone resellers will start advertising phone installment plans which are not tied to the service plan. Consumers could see the real costs of the phone vs. the service, choose to pay for the phone up front or on installment, and get out of their service plan any time without losing anything. How would they be worse off?
It's very easy when your audience is filled with people who do not understand distinctions such as "could be seen as racist" vs "racist". Here's a page discussing it:
The movie ends when Moore walks up to Charlton Heston's home and just rings the buzzer. Heston answers and invites Moore back for an interview the next day. At the interview, the then 77-year-old Heston holds his own, but Moore's relentless attack makes the legendary screen actor and civil rights activist seem befuddled and confused.
Moore continues to pester Heston as to why Canada has a lower crime rate in proportion to their large gun ownership rate; Heston then comments on our mixed ethnicity. Moore asks him if he thought race was the cause of crime in America, and Heston flatly says no. However, many reviewers have picked that one sentence out to accuse Heston of being a racist. Heston may be many things but a racist is not one of them.
For once, I'd like one of Moore's critics to address the points he makes rather than the techniques he uses to make them.
That's a bit hard to do because his most insidious, dishonest points are ones he denies making at all. His subtle edits and omissions of facts lead the viewer to an erroneous conclusion that was never explicitly stated.
It's been a long time since I looked into this, but iirc Moore manipulated Charlton Heston into saying something which could be seen as racist (neglecting to mention that the man actually marched with MLK Jr!), implied that Heston led a rally in Colorado in response to Columbine (it was planned well in advance), and that Heston used language which, given the situation, was quite inflammatory (those snippets were actually spliced together from several different rallies). He made many other, less topical, manipulations of fact.
I suspect it has something to do with his politics, since the ones who complain the loudest tend to be die-hard Republicans.
I am a Democrat, and I am disappointed that so few Democrats complain about Moore.
Damn, I'm totally independant, but I wish I was republican for a moment, if for nothing else but to throw out that zinger.
No need to be Republican - it's important to admit when people on your side are being dishonest. (And "Bowling for Columbine" misrepresented Charlton Heston's actions and basically everything else it portrayed.) One could argue that the party of Lincoln fell so far because Republicans wouldn't call out Republicans for ethical lapses, and it'd be sad if the Democrats followed them any further down that road.
The question was: "With knowing the platform, how could you say?" That's what odoketa is asking you. To reiterate, if he had given you the name of the platform, by what process would you decide if he should ditch it?
Everyone here (including you, obviously) seems to be expecting him to just ask a expert if his specific platform should be ditched, and I'm inclined to agree. So my answer to his question is "ask an expert, preferably one who works for you". If you're not going to invest the effort to become an expert, there aren't any real shortcuts to making this kind of decision, so you need to hand it off to someone who has invested that effort.
So here's my new question: "How can a nontechnical person hire a good web developer to take over a project written in platform X?"
odoketa's current approach seems to be finding web developers who know platform X. That's not working well. There are a lot of platforms out there, and given the proportion of good web developers to all web developers, restricting yourself further to ones who happen to know a particular platform seems unreasonable. If they're really good, they can learn it quickly, particularly if they've used similar platforms in a similar programming language before.
Unfortunately, I don't have a good idea. I'm a technical person, so I have the luxury of being able to ask candidates a variety of technical questions to determine if they know what they're talking about. That only works if you already have at least one technical person on hand...
Oh, an alternate solution - if you never send too much interactive traffic, you could just limit your torrent's upload speed to be a bit less than the link speed. Again, this ensures the cable modem's ridiculously long queue never fills. rtorrent can do this in userspace from the torrent UI.
Unfortunately, I recently moved back under comcast's umbrella and had no other options for internet within my budget. And now I'm suffering latencies as high as three seconds whenever I download a torrent. As soon as I stop torrent downloads, my latency returns to 25ms.
This is industry-standard practice (it didn't happen with your last ISP? really? it happens with AT&T and I've heard people with many other ISPs complain of it as well), and you can fix this on your side.
The problem is simple. They have ridiculously long queues (both for you->them on the cable modem and them->you on their equipment) to optimize for throughput, not latency. As you say, it isn't really "traffic shaping", although they may be doing QoS as well. I've verified that my solution below completely solves the problem on AT&T's network. Though I'm a Comcast customer, I haven't tried it on Comcast's network yet.
The solution? First, recognize that the more problematic leg is likely you->them - BitTorrent will suck up both, and you have less upload capacity than download capacity. (This is fortunate, because the other problem is not as easy for you to solve.) Traffic shaping is in fact the solution, not the problem. Ensure your interactive packets go into a short, high-priority queue: the queue's maximum contribution to the latency will be the total time to sent its contents. Also ensure you don't drop interactive packets if possible, so put the BitTorrent traffic into a lower-priority queue. For BitTorrent packets, you care more about throughput than latency, so it makes sense to have that queue be a bit longer. And finally, ensure you never fill the cable modem's much longer queue by limiting your sending rate to the link speed (tc-tbf can do this). You say you're familiar with tc, so you should have no problem creating this arrangement.
To maintain that immunity, AT&T must transmit data 'without selection of the material by the service provider' and 'without modification of its content'
Well, neither of the criteria contains any mention of the transfer rate. They could limit "offending" downloads to 1 kB/s.
Not really, no. I think you're legally shaky saying that ("without selection of the transfer rate" does not mean "you can transmit it differently, selecting by the transfer rate, but you have to transfer it") but I'm not a lawyer, so instead I will tell you why your approach eventually comes down to dropping packets.
As you may know, QoS works by having different queues (potentially of varying lengths), enqueueing policy (like add X packets to queue X, unless it's full, then queue Y, unless it's full, then drop), and dequeueing policy (like pull from queue X unless that would exceed rate Y, queue Z otherwise).
Say they do what you suggested - specifically have an "offending download" queue, only dequeue from it every 8 kbps, and never put those packets in any other queue. If I send them "offending download" packets at 8 Mbps, they're forced to create an arbitrarily long queue or drop some of my packets. (I can make this possible by using large TCP windows or simply using a UDP protocol with no flow control at all.) They must queue terabytes of my information (and potentially terabytes for everyone else as well; totally infeasible), give up on their queueing scheme, or drop "offending downloads" packets completely out of proportion to their other queues.
This is possible even without any effort on my part - at the rates you suggested, with a number of people trying, they will have to drop a huge number of packets for people using standard window sizes.
I particularly like the way they've photographed an SD card and GIMPed it into their 'product' but not mentioned anything about it being concept art or anything too.
I attended CES 2005 and saw their booth. You're wrong; this is not a Photoshop/GIMP job. They actually took SD cards and physically stuck things (labels and optical things) on them. I was tempted to ask for a demonstration, but I was afraid to let them plug anything into my laptop.
OK, they need to get the price down to $100. Instead of selling them in the US at $400 at a 100% profit margin to raise money for charity, they need to just sell the things for $199 commercially and take over the low end market. In a year or two, they'll be down to $99 through sheer volume.
$100 is in some ways a harder target now than when they originally set it. I don't recall exactly when they first talked of it costing $100, but the dollar has fallen in the last five years (e.g. 25% against the GBP), and the OLPC folks have upgraded their system specs (e.g., 128 MB RAM -> 256 and 512 MB flash -> 1024).
Those things ought to be in bubble-packs at the local drugstore, alongside the cheap calculators, electronic dictionaries, and other low end electronics. This wouldn't stroke Negroponte's ego, but it would get the things out there in volume. Soon enough, they'd be available all over the world, purely on price.
They're doing pretty well on volume now. They have a brand-new factory, and last month they planned to ship 150,000, then 80,000-100,000 every month after (source).
Where are they going? I just did a bit of hunting. Uruguay ordered 100,000 units(see wiki) and Peru ordered 260,000 (see this post, near bottom). According to the "country news" section, Mexico's also placed some order; I think 100,000 is the minimum order size. 150,000 to 170,000 individual G1G1 orders and 15,000 for Birmingham, Alabama, for a total of around 400,000 G1G1 laptops (see interview), so I believe they have solid orders for 800,000 laptops.
Hopefully when they've had success with those 800,000, the other countries that originally intended to be part of the launch will get back on the bandwagon. So while I'm not a manufacturing expert, I would guess the difference between 1 million/year and 2 million/year isn't going to hugely affect the cost.
In addition to the problems you mentioned, I'm worried by the fact that they don't describe in detail what they mean by "placebo." For instance, they mention "two separate rooms" in their experimental section, but don't explain why they have two rooms; if one was "real" and the other "placebo" then the variability could easily be ascribed to minor variations in the rooms (lighting, ambient sound, odor, etc.). The RF transmitter is placed immediately beside the person's head (there is a photo in the article), which worries me because they never mention measuring or accounting for audio effects: a high-pitched whine from a running device could easily explain the differences (it wouldn't even have to be consciously audible to influence the subjects).
Interesting point about the audio, and I imagine they would have mentioned it if they'd checked for that, so...
Regarding the rooms: you would know: is it normal for a study of this type for them to show only statistical results? My understanding was that they would make available their complete raw data for verification, including any other variables like which room the subject was in.
They do say that the experiment was double-blind. One would hope then that they don't have a "real" room and "placebo" room, as it couldn't possibly be double-blind if that were true. But I do wonder if it was fully double-blind, because they don't say how they achieved that. There was probably a person who placed the apparatus on their heads and later removed it, handed them the study, then hooked them up to the EEG? Could that person have discovered if the unit was on?
Combined with the very large standard-deviations on their results, I'm hesitant to ascribe any significance to this finding just yet. More details, and corroborating independent verification, are definitely necessary before raising any public alarms.
Their p-values are quite low, though (p=0.0037 and p=0.0019)? Isn't that good enough?
On the other hand:
The strengths of this study compared to earlier studies are
the longer exposure time during worst conditions and a wider range of outcome variables, including
self-reported, neuroendocrine, and neurophysiological variables.
I think this means they gathered a lot of data (much more than length to stage-{3,4} sleep) and headaches), made a lot of hypotheses, and only presented the fraction that were statistically significant.
You are simply never going to protect all the stupid people from themselves, and making the effort often only punishes the smart people who didn't make those mistakes. That's the unfortunate realization I've come to in my adulthood.
In this case, stupid people are not the only ones in danger from their stupidity. I say that not only because stupid people's children are not necessarily stupid, but also because vaccines are not 100% effective. Their benefit to an individual comes not only from that individual being vaccinated (lowering chances of catching the disease if exposed to it) but also from everyone else being vaccinated (lowering chances of other people catching it => lowering chances of being exposed to it). If there is a significant population who have not received the vaccine (a.k.a. idiots' children), everyone is at greatly increased risk. That is why governments and schools make some vaccines mandatory - non-immunized people are a threat to the common good.
Sounds like a great opportunity for the law enforcement officers of Barrow Gurney to make some money issuing fines.
To me, it sounds like a rare instance of authorities caring more about safety than money. Unfortunately, your attitude seems to be more common - to the point that some communities (*cough*Union City, CA*cough) have been caught deliberately and illegally causing unsafe situations in order to increase revenue from traffic violations.
Sure doesn't, but even if they had, ftok() is a lousy excuse for a hash function. From the Darwin ftok() manpage, BUGS section:
The returned key is computed based on the device minor number and inode of the specified path in combination with the lower 8 bits of the given id. Thus it is quite possible for the routine to return
duplicate keys.
IIRC from reading Stevens, way back when the interface was designed 32 bits was sufficient for guaranteed uniqueness. Even now, you could achieve somewhat low probability of collision with a decent hash function. But ftok() is not a decent hash function, so it's probably no better than everyone just picking numbers arbitrarily. It's a totally braindead interface now that mmap() exists on all useful platforms.
mmap() has plenty of other advantages, like actual reference counting (open descriptors + filesystem links) instead of a virtual post-it note reading "last person to leave the shared memory segment please turn the lights off with shmctl(..., IPC_RMID,...)".
With SystemV shared memory (shmem) it's trivial, and that's a decades-old feature of Real Unixes. What, doesn't OSX support it?
What? Of course it does. It also supports the vastly superior mmap(), though unfortunately not everyone is using it. I remember struggling one Christmas morning to get iChat AV to work because iChat and PostgreSQL were fighting over the same shmget() key_t (54321, IIRC). mmap() allows you to just use a filename, which is much easier to make unique.
The whole tone of the article can be summed up here:
About 4,000 children were afflicted with cancer. Less well-known, however, is the fact that only nine of those 4,000 died -- thyroid cancers are often easy to operate on.
See there, not so bad! "Only" nine people died. The 3991 others did not mind having their thyroid glands removed at all. All is well that ends in useless pain and suffering.
This article makes me sick.
The whole tone of your post can be summed up here: "The 3991 others did not mind having their thyroid glands removed at all." Except...they really didn't say that. I believe the most they said was that it was better than dying, and that most people do not know that those children did not die. I agree with both points.
You haven't disputed any actual claims of the article. So why are you opposing the harmless gathering of information for scientific study and the presentation of surprising but true information? You speak as if the authors personally caused the children to have thyroid cancer just to see how many of them would die, or as if they are saying nuclear waste is totally harmless and advocating fertilizing our crops with it. I don't like the idea of censoring science because you apparently find the results to be politically inconvenient.
Slashdot Mirror
You're obviously speculating wildly without doing your homework. Your claims are at best unsupported, and your claim that they don't want to pay for the testing is directly contradicted in the blog entry and legal brief in the following respective paragraphs:
You filed at least two P1 bugs right?
and mentioned that they were definitely costing the company $X,XXX/day in support and probably costing more in lost customers as bad word-of-mouth spread?
Arguably the others - broken installer system (Microsoft's bug), bad hardware, and pre-install virus (which may not have existed when the media shipped) are not defects in your product, but they're costing you money as well, so they should be worked around if at all possible.
It'd be easy to write a small program that requires little RAM and no disk (beyond the binary) which would automate at least the "delete the files and do this in regedit" portion of your instructions to grandma. Ship it with the media and (for customers with Internet connections and installations predating the fix) put it on your website. Call it "uninstaller version N+1".
Working around serious software defects should not be considered an "ordinary" task.
I hope you're saying that because he's somewhat young and the crimes were (presumably) his first and non-violent, not because most were computer crimes (a "clever" white-collar crime many here could identify with). Why not apply the same reasoning to, say, a high school student who commits a string of burglaries/thefts?
All right. It's a work in progress (no keepalives yet; requires libevent trunk + a few not-yet-merged patches), but I'm willing to share what I have. I've traveling now, but I'll put something on my website Monday night. I'll call it FoxIgniter, unless someone has a better name.
I wrote a proxy server to run on localhost, do DNS resolution, and send the request to the appropriate upstream proxy or directly to the source, just to work around this bug.
Many Democrats would now agree with you and the Republicans that Hillary Clinton is disingenuous and power-hungry.
Certainly she was disingenuous in her claims of winning the popular vote. She included the votes in Michigan, where Obama followed rules by withdrawing his name but she did not. She supposedly did this to because every vote should count, but her statements showed no regard for that before she fell far behind. At the same time, she's certainly not counting every vote by arbitrarily excluding all caucus states from consideration.
Power-hungry in that her actions have led us to fear she'll tear the party apart and give ammunition to the Republicans rather than admit she's lost. Again with Michigan. Her lawyers' statements make me fear she'll launch a legal challenge to the assignment of Michigan's "uncommitted" delegates to Obama. As irregular as it was, the only better way to represent the wishes of those voters would be to have a new vote. I imagine she'd oppose that because Obama would win it. (Even ignoring consideration of the events between January 15th and now, I think it's fair to say that if Obama's name were on the ballot, more people would have shown up to vote for him than actually did show up to vote for "uncommitted". Given that 40% of the people who did show up voted for "uncommitted" or "undecided", that likely means he would have won.)
I'd also throw in "pandering", particularly in her suggestions of foreclosure and gas tax holidays. ("I'm not going to put my lot in with economists.") The only other explanation I see (that's she's not only missing the obvious flaws in these schemes but also rejects the concept of people who devote their professional lives to studying the issues having a better understanding of them) is not a good quality either.
I had a sliver of doubt when I cast my vote for Obama instead of Clinton in California's Democratic primary. That sliver is long since gone.
We should vote for someone else because other people will vote for someone else because his name reminds them of two bad people?
I want to live in a world where people don't cast their votes for such stupid reasons. And to quote Mahatma Gandhi, "you must be the change you want to see in the world." If that makes me too idealistic, so be it. Better that than too cynical like you.
Right; I'm not willing to spend several hundred dollars in a lump sum up front. That's because I'm paying hidden monthly fees for the phone already. If I break the contract, I'll pay an early termination fee and be left with a useless service-locked phone, and my next contract will inevitably pay for a new phone anyway. So what would I gain by paying money up front, too?
Striking down cancellation fees and allowing open access (any phone on any network) would change the game. I'd gladly pay for my phone outright.
But let's assume you're right - even without those gotchas, most people would want the most expensive phone on an installment plan. If the carrier early termination fees go away, cell phone resellers will start advertising phone installment plans which are not tied to the service plan. Consumers could see the real costs of the phone vs. the service, choose to pay for the phone up front or on installment, and get out of their service plan any time without losing anything. How would they be worse off?
That's a bit hard to do because his most insidious, dishonest points are ones he denies making at all. His subtle edits and omissions of facts lead the viewer to an erroneous conclusion that was never explicitly stated.
It's been a long time since I looked into this, but iirc Moore manipulated Charlton Heston into saying something which could be seen as racist (neglecting to mention that the man actually marched with MLK Jr!), implied that Heston led a rally in Colorado in response to Columbine (it was planned well in advance), and that Heston used language which, given the situation, was quite inflammatory (those snippets were actually spliced together from several different rallies). He made many other, less topical, manipulations of fact.
I am a Democrat, and I am disappointed that so few Democrats complain about Moore.
No need to be Republican - it's important to admit when people on your side are being dishonest. (And "Bowling for Columbine" misrepresented Charlton Heston's actions and basically everything else it portrayed.) One could argue that the party of Lincoln fell so far because Republicans wouldn't call out Republicans for ethical lapses, and it'd be sad if the Democrats followed them any further down that road.
The question was: "With knowing the platform, how could you say?" That's what odoketa is asking you. To reiterate, if he had given you the name of the platform, by what process would you decide if he should ditch it?
Everyone here (including you, obviously) seems to be expecting him to just ask a expert if his specific platform should be ditched, and I'm inclined to agree. So my answer to his question is "ask an expert, preferably one who works for you". If you're not going to invest the effort to become an expert, there aren't any real shortcuts to making this kind of decision, so you need to hand it off to someone who has invested that effort.
So here's my new question: "How can a nontechnical person hire a good web developer to take over a project written in platform X?"
odoketa's current approach seems to be finding web developers who know platform X. That's not working well. There are a lot of platforms out there, and given the proportion of good web developers to all web developers, restricting yourself further to ones who happen to know a particular platform seems unreasonable. If they're really good, they can learn it quickly, particularly if they've used similar platforms in a similar programming language before.
Unfortunately, I don't have a good idea. I'm a technical person, so I have the luxury of being able to ask candidates a variety of technical questions to determine if they know what they're talking about. That only works if you already have at least one technical person on hand...
Oh, an alternate solution - if you never send too much interactive traffic, you could just limit your torrent's upload speed to be a bit less than the link speed. Again, this ensures the cable modem's ridiculously long queue never fills. rtorrent can do this in userspace from the torrent UI.
This is industry-standard practice (it didn't happen with your last ISP? really? it happens with AT&T and I've heard people with many other ISPs complain of it as well), and you can fix this on your side.
The problem is simple. They have ridiculously long queues (both for you->them on the cable modem and them->you on their equipment) to optimize for throughput, not latency. As you say, it isn't really "traffic shaping", although they may be doing QoS as well. I've verified that my solution below completely solves the problem on AT&T's network. Though I'm a Comcast customer, I haven't tried it on Comcast's network yet.
The solution? First, recognize that the more problematic leg is likely you->them - BitTorrent will suck up both, and you have less upload capacity than download capacity. (This is fortunate, because the other problem is not as easy for you to solve.) Traffic shaping is in fact the solution, not the problem. Ensure your interactive packets go into a short, high-priority queue: the queue's maximum contribution to the latency will be the total time to sent its contents. Also ensure you don't drop interactive packets if possible, so put the BitTorrent traffic into a lower-priority queue. For BitTorrent packets, you care more about throughput than latency, so it makes sense to have that queue be a bit longer. And finally, ensure you never fill the cable modem's much longer queue by limiting your sending rate to the link speed (tc-tbf can do this). You say you're familiar with tc, so you should have no problem creating this arrangement.
Not really, no. I think you're legally shaky saying that ("without selection of the transfer rate" does not mean "you can transmit it differently, selecting by the transfer rate, but you have to transfer it") but I'm not a lawyer, so instead I will tell you why your approach eventually comes down to dropping packets.
As you may know, QoS works by having different queues (potentially of varying lengths), enqueueing policy (like add X packets to queue X, unless it's full, then queue Y, unless it's full, then drop), and dequeueing policy (like pull from queue X unless that would exceed rate Y, queue Z otherwise).
Say they do what you suggested - specifically have an "offending download" queue, only dequeue from it every 8 kbps, and never put those packets in any other queue. If I send them "offending download" packets at 8 Mbps, they're forced to create an arbitrarily long queue or drop some of my packets. (I can make this possible by using large TCP windows or simply using a UDP protocol with no flow control at all.) They must queue terabytes of my information (and potentially terabytes for everyone else as well; totally infeasible), give up on their queueing scheme, or drop "offending downloads" packets completely out of proportion to their other queues.
This is possible even without any effort on my part - at the rates you suggested, with a number of people trying, they will have to drop a huge number of packets for people using standard window sizes.
(BTW, I misspoke; I was at CES 2006, not 2005.)
I attended CES 2005 and saw their booth. You're wrong; this is not a Photoshop/GIMP job. They actually took SD cards and physically stuck things (labels and optical things) on them. I was tempted to ask for a demonstration, but I was afraid to let them plug anything into my laptop.
$100 is in some ways a harder target now than when they originally set it. I don't recall exactly when they first talked of it costing $100, but the dollar has fallen in the last five years (e.g. 25% against the GBP), and the OLPC folks have upgraded their system specs (e.g., 128 MB RAM -> 256 and 512 MB flash -> 1024).
They're doing pretty well on volume now. They have a brand-new factory, and last month they planned to ship 150,000, then 80,000-100,000 every month after (source).
Where are they going? I just did a bit of hunting. Uruguay ordered 100,000 units(see wiki) and Peru ordered 260,000 (see this post, near bottom). According to the "country news" section, Mexico's also placed some order; I think 100,000 is the minimum order size. 150,000 to 170,000 individual G1G1 orders and 15,000 for Birmingham, Alabama, for a total of around 400,000 G1G1 laptops (see interview), so I believe they have solid orders for 800,000 laptops.
Hopefully when they've had success with those 800,000, the other countries that originally intended to be part of the launch will get back on the bandwagon. So while I'm not a manufacturing expert, I would guess the difference between 1 million/year and 2 million/year isn't going to hugely affect the cost.
Interesting point about the audio, and I imagine they would have mentioned it if they'd checked for that, so...
Regarding the rooms: you would know: is it normal for a study of this type for them to show only statistical results? My understanding was that they would make available their complete raw data for verification, including any other variables like which room the subject was in.
They do say that the experiment was double-blind. One would hope then that they don't have a "real" room and "placebo" room, as it couldn't possibly be double-blind if that were true. But I do wonder if it was fully double-blind, because they don't say how they achieved that. There was probably a person who placed the apparatus on their heads and later removed it, handed them the study, then hooked them up to the EEG? Could that person have discovered if the unit was on?
Their p-values are quite low, though (p=0.0037 and p=0.0019)? Isn't that good enough?
On the other hand:
I think this means they gathered a lot of data (much more than length to stage-{3,4} sleep) and headaches), made a lot of hypotheses, and only presented the fraction that were statistically significant.
In this case, stupid people are not the only ones in danger from their stupidity. I say that not only because stupid people's children are not necessarily stupid, but also because vaccines are not 100% effective. Their benefit to an individual comes not only from that individual being vaccinated (lowering chances of catching the disease if exposed to it) but also from everyone else being vaccinated (lowering chances of other people catching it => lowering chances of being exposed to it). If there is a significant population who have not received the vaccine (a.k.a. idiots' children), everyone is at greatly increased risk. That is why governments and schools make some vaccines mandatory - non-immunized people are a threat to the common good.
To me, it sounds like a rare instance of authorities caring more about safety than money. Unfortunately, your attitude seems to be more common - to the point that some communities (*cough*Union City, CA*cough) have been caught deliberately and illegally causing unsafe situations in order to increase revenue from traffic violations.
Sure doesn't, but even if they had, ftok() is a lousy excuse for a hash function. From the Darwin ftok() manpage, BUGS section:
IIRC from reading Stevens, way back when the interface was designed 32 bits was sufficient for guaranteed uniqueness. Even now, you could achieve somewhat low probability of collision with a decent hash function. But ftok() is not a decent hash function, so it's probably no better than everyone just picking numbers arbitrarily. It's a totally braindead interface now that mmap() exists on all useful platforms.
mmap() has plenty of other advantages, like actual reference counting (open descriptors + filesystem links) instead of a virtual post-it note reading "last person to leave the shared memory segment please turn the lights off with shmctl(..., IPC_RMID, ...)".
What? Of course it does. It also supports the vastly superior mmap(), though unfortunately not everyone is using it. I remember struggling one Christmas morning to get iChat AV to work because iChat and PostgreSQL were fighting over the same shmget() key_t (54321, IIRC). mmap() allows you to just use a filename, which is much easier to make unique.
The whole tone of your post can be summed up here: "The 3991 others did not mind having their thyroid glands removed at all." Except...they really didn't say that. I believe the most they said was that it was better than dying, and that most people do not know that those children did not die. I agree with both points.
You haven't disputed any actual claims of the article. So why are you opposing the harmless gathering of information for scientific study and the presentation of surprising but true information? You speak as if the authors personally caused the children to have thyroid cancer just to see how many of them would die, or as if they are saying nuclear waste is totally harmless and advocating fertilizing our crops with it. I don't like the idea of censoring science because you apparently find the results to be politically inconvenient.
Your post makes me sick.