Nor do I say otherwise. I was merely demonstrating my willingness to avoid x86.
Yeah, loongson's not exactly floodying the market with earthshattering specs and commodity pricing. However, for me, http://www.tekmote.nl/Loongson-3A-Notebook would be an upgrade laptop-wise. (But I wouldn't be willing to stretch to even that reduced price.) However, at my age, and with my interests, my main interest would be smaller-quieter-leaner rather than more raw grunt. I may never buy an actual "desktop" machine again. Rendering a dozen or so xterms is all I need most of the time.
Ah, I'm certainly *way* more familiar with central london than anywhere else, but the place I bought in the midlands was leasehold too. As was at least one of the places my sister owned up north. Hasty generalisation on my part. If the "But thanks to changes made over the past decade, they have the right to get together with their neighbours and buy the landlord out" in that article has had much effect it may have skewed things as I've not lived in the UK since the 90s, that that would surprise me. Thanks for the numbers.
OK, the article itself is not wrong. I just didn't want people to jump to the conclusion that because a modern HDD says "Samsung" on it, it's not a US device. (In the context of "Don't buy from US companies", i.e. this sub-thread.)
It was unfortunate that the article mentioned the US-iness of those manufacturers, so I conflated the two sentences and caused confusion.
Alas it's rather hard to avoid intel & AMD for those who are tied to the architecture.
Now my Alpha's long dead, and my POWER is getting rather long in the tooth, I suspect my next purchase will be a Loongsoon-powered box.
Note that the mention of Samsung in the article is a little bit wrong - they sold their HDD division to Seagate (a US company, modulo tax-evasion) in 2011. Whether their SSDs are compromis{ed,able} is another matter.
My post was supposed to be read in the context of using focussing on metamaterial approaches. I.e. what this story is about. If you forget the context of the story, then my post makes less sense. The problem has always been scale, as metamaterials work at the scale of the wave, and the scale of these waves is mind-bogglingly huge. Therefore it's inappropriate to try and process the wave behaviour in order to solve the problem of the bulk matter.
Your completely backwards about why tidal waves were named that way. They were named that way as they were tide-like in effect. The name is appropriate. If you start saying "but they're not caused by tides", then you deserve the response "nobody said they were, fool". If you stop deliberately misinterpretting things, you'll have a lot less stressful life. You probably think that my weight should only be measured in Newtons rather than kilograms, don't you? If so, yet again, you'd be wrong.
Maybe only 17 queries are required. So even if they did to some kind of rate-limitting to prevent mass sucks of account names, they'd not stop the leak.
Number all the names you're interested in binary. If you have 75000 names, then the binary numbers will be 17 bits long. In the first query, do a lookup on the (75000-65536) contacts which have a set 16th bit. Store all the results. In the second query, do a lookup on all the 32768 contacts which have a set 15th bit, again, store those. In the third query, do a lookup on all the (16384+16384) contacts which have a set 14th bit, again store. After 17 queries, each contact will be returned in exactly the sets which correspond to the bits that are set in its binary number, but not the others. I.e. it will be uniquely identifiable.
Of course, the fix for the problem is for the doofera at snapchat to simply not return account names in the query, and this 4000x speedup will stop working as quickly as the original. However, anyone who's done a huge suck prior to that could leak it out, so it must be considered that your account name is known to everyone. Expect more targetted adverspamming...
We are told (QI?) that US public phone booths do not have doors because a door would create a reasonable expectation of privacy. The implication is that if since that decision was made, you make a phone call from such a booth, all information about it is fair game to whomever. So your use of telecoms facilities has never been particularly private. (Personally, from that superficial summary above I think even Smith v. Maryland has gone too far. Why was there no warrant? Would that have required too much police-work? Like the "hunch"es that we see on the movies?)
I think the "how on earth can SCOTUS get away with that?!?!?" moment was when an ex-Monsanto lawyer thereon didn't declare an interest and then judged in favour of Monsanto.
Sure, they've been way more wrong (cough, Dred Scott, cough), but at least that was a clear and open wrongness (in that case, some of the most racist frothing I've ever read someone in a state of power utter), rather than a corrupt ulterior one.
Just outside the US, to the south, you'll find the United Mexican States.
Time for a classic: The United Nations initiated a poll with the request, "Please tell us your honest opinion about the lack of food in the rest of the world." The poll was a total failure. The Russians did not understand "Please". The Italians did not know the word "honest". The Chinese did not know what an "opinion" was. The Europeans did not know "lack", while the Africans did not know "food". Finally, the Americans didn't know anything about the "rest of the world".
Do you not have housing associations? Where you "own", but you own a share in the association which grants you the right to live in the property that your share corresponds to.
Likewise, do you not have leasehold ownership, rather than freehold ownership? In the UK, almost everything that's "owned" is leasehold. (You own the building and the right to live there, but not the land its on.)
Do you also not have enough soft recycled wood pulp to use as toilet paper? Otherwise, why are all three branches of government using the constitution for that task?
And if you throw 10000 notes at the problem, you can take a photo where both the eye itself, which takes up a large proportion of the frame it is so close and the reflected scene about 1-2m away are in perfect focus, and with vivid colours to boot. One of the best photos I've seen in the last few years, and when my mate says "no photoshop apart from when gamut-mapping from raw to jpeg", I believe him. I'm still not sure I know exactly how he did it so brilliantly, I'm guessing it was very careful lighting.
Sounds like you know more about what actually happens in the US than either me or target:
""" Target does not have access to nor does it store the encryption key within our system. The PIN information is encrypted within Targetâ€(TM)s systems and can only be decrypted when it is received by our external, independent payment processor. """ -- http://pressroom.target.com/news/target-data-security-media-update-4
That first "encryption key" should read "decryption key", surely? How can they perform encryption within their system without the encryption key being stored somewhere within their system?
The whole system makes no sense to me. The chip on the card should verify the PIN. It should never be transmitted out of the terminal, definitely not sent to any third party, and it should certainly never be stored anywhere. Once the chip has verified the PIN, that yes/no answer is all you need to store. At least that's how things work in backwards eastern Europe.
Is there any reason to believe that whatever is required to get from the super-secret encryption key to the session encryption key (a terminal identifier, timestamp, nonce?) has not also been leaked along with the encrypted PIN and other data? Is the distinction between an "encryption" and a "decryption" key meaningful (is this symmetric crypto, the above quote is so mangled it's hard to trust anything in it?) - as even in the absense of a decryption key, an encryption key would permit the creation of a dictionary (or equivalenly a 10^4/10^6 brute force) for finding which PIN encrypted to the particular cyphertext.
Now reread the paragraph you're responding to. It's hard for you to be more wrong. It's not a reference point, *meaningless*. People who work with databases will never have *any* interest in this toy. Ever. As someone currently looking at petabyte databases, I even laugh at the few-terabyte setups now. A tweak I was looking at this month managed to save a few terabytes in storage for one typical scenario. It's uncertain if that's significant enough to be worth implementing - more research required.
> tell me where on the ROI list partnering on academic research papers falls. It's not there
False. That's basically the entire raison d'etre for the creation of the last place where I worked, a supposed "R&D" centre. It's part of what my job was. So I'm speaking from experience of the cutting-edge IT field in 2013. They were happy to throw a good 7 figures in that direction. Feel free to ask them where the ROI was, all I know is that it happens because it happened.
> Have an interesting book buying list, travel [...]
Real world example I know of personally: Have a (nominally Christian) boyfriend from a country where the prevailing religion is Islam. Bang - straight onto the list. So much so that the pair in question even picked up a tail of spooks at least once when on holiday.
You're confusing "actually tides" and "like tides". The closest thing to a tsunami is a tide. Tsunamis are about as tidal as you can get without actually being tides.
Yes, you can model the uninterrupted propagation of tsunamis as waves. Yes, you can model the diffraction around island-sided opjects as waves. Yes, you can model the propagation of tsunamis down channels as waves, solitons as you say. But no, humans cannot realisitically use meta-material approaches to modifying tsunamis in order to protect against them.
Unless you can build *island-sized* features. The wavelenth is 500-1000km. The meta-material needs features with a size comparable to that of the wavelenth in order to affect wave-like behaviour in the ways being talked about. We do not have the capability to build such structures. I've not read the particular research the article refers to, but I presume it's to do with the negative refractive index lens. That permits you to create a shadow (i.e. shield what is behind) which is larger than the object itself. (So in theory, small things could protect an island.) However, in order to do that, you have to build a metamaterial structure the size of the shadow you wish to create. That's not a win. It's a massive loss, as now you have to build a huge fancy thing rather than a huge dumb thing.
They've also recruited at least 20 people whose response to "and what was your password?" is their password. It doesn't matter how many "brightest minds" you have if you have such weak links.
They did not publish a formal TPC-H result, they simply name-dropped TPC-H. In a way that, as it can't be compared to anything, TPC themselves is "meaningless".
And the 3 guys with nvidia email addresses on the paper have no interest in how well nvidia cards sell? It is you who seems confused, or at least naive. This was nvidia marketting. The company I just quit from used to do exactly the same - they all do.
Unless those processors are trying to use the same bus, in which case, more processors can mean more contention. Worst of all - by a superlinear amount.
Don't try to optimise until you've measured what you are most often doing with the box, in particular what you are most often waiting for. If you're very lucky, you might be compute bound, in which case, the more the merrier when it comes to CPUs (but only if the software is coded to scale sensibly). However, in my real-world experience (rather than hobby stuff), being I/O-bound is far more common.
The greatest amount of damage from tsunamis happens when *other things* bend the tsunami. Anything which diverts the energy away from one thing diverts it towards another thing. Heaven help you if you've got two things diverting the energy towards you, such as sides of a bay, as then you're more likely to have an e^-kd rather than k.d^-2 reduction in energy (an extreme - compare losses in fibre optics to the inverse square law).
Thanks for the correction. I did check and realise my error after posting, but felt it unimportant, as the way the benchmarks had appeared, the H on was the one used on the at-the-time huger datasets. As time has progressed, they (TPC) now just say it's for "large" volumes of data, but they do clearly emphasise that the size is relevant. Maybe I picked up the folk meaning from a journo, and didn't bother verifying.
But anyway, even if they clustered 8 nodes together, which they didn't do, 96GB would still not cut it. The whole point of TPC was for real-world scenarios. These guys are still playing with toys. This is not a breakthrough, this is merely an interesting line of investigation. It might lead to a breakthrough, certainly, but they should only publish results when they're actually playing in the same ball-park, not 3 orders of magnitude away.
Yeah, but the solution to the problem in high energy physics is to make sure the high energy things don't touch anything important.
So, in this case, all we need to do is to levitate the ocean.
And you do realise that the mass of these pulses that LHC is trying to bend is only a tenth of a millionth of a gram? Tsunamis weigh quite a bit more than that.
But jesting aside, the article does look mostly bullshit, as it's highly inappropriate to model tsunamis as waves and solutions as working upon waves. They're so low frequency they're effectively DC. That's why in the term "tidal wave" is inappropriate - the "tidal" part is fine - in fact, it's almost perfect - it's the "wave" part that's misleading.
There's a weapon that works against such behaviour - being patronising. It only works slowly, but it does work eventually. Don't snark here - just laugh "useless use of pipes", or "useless use of cat", or "useless use of $whichever_thing_he_used_which_was_useless" every time you see it. It's a hundred times easier and quicker to laugh than it is to type a dumb command line, so it's a very cheap remedy.
I had a very bright co-worker who had never encountered the '-w' switch in grep. It took precisely one respectfully-made snark to fix that behaviour. He's optimised a few of the tasks that I commonly do too. Those who care learn, those who don't care get laughed at.
Slashdot Mirror
> Alpha was american too, as is POWER...
Nor do I say otherwise. I was merely demonstrating my willingness to avoid x86.
Yeah, loongson's not exactly floodying the market with earthshattering specs and commodity pricing. However, for me, http://www.tekmote.nl/Loongson-3A-Notebook would be an upgrade laptop-wise. (But I wouldn't be willing to stretch to even that reduced price.) However, at my age, and with my interests, my main interest would be smaller-quieter-leaner rather than more raw grunt. I may never buy an actual "desktop" machine again. Rendering a dozen or so xterms is all I need most of the time.
Ah, I'm certainly *way* more familiar with central london than anywhere else, but the place I bought in the midlands was leasehold too. As was at least one of the places my sister owned up north. Hasty generalisation on my part. If the "But thanks to changes made over the past decade, they have the right to get together with their neighbours and buy the landlord out" in that article has had much effect it may have skewed things as I've not lived in the UK since the 90s, that that would surprise me. Thanks for the numbers.
OK, the article itself is not wrong. I just didn't want people to jump to the conclusion that because a modern HDD says "Samsung" on it, it's not a US device. (In the context of "Don't buy from US companies", i.e. this sub-thread.)
It was unfortunate that the article mentioned the US-iness of those manufacturers, so I conflated the two sentences and caused confusion.
Alas it's rather hard to avoid intel & AMD for those who are tied to the architecture.
Now my Alpha's long dead, and my POWER is getting rather long in the tooth, I suspect my next purchase will be a Loongsoon-powered box.
Note that the mention of Samsung in the article is a little bit wrong - they sold their HDD division to Seagate (a US company, modulo tax-evasion) in 2011. Whether their SSDs are compromis{ed,able} is another matter.
My post was supposed to be read in the context of using focussing on metamaterial approaches. I.e. what this story is about. If you forget the context of the story, then my post makes less sense. The problem has always been scale, as metamaterials work at the scale of the wave, and the scale of these waves is mind-bogglingly huge. Therefore it's inappropriate to try and process the wave behaviour in order to solve the problem of the bulk matter.
Your completely backwards about why tidal waves were named that way. They were named that way as they were tide-like in effect. The name is appropriate. If you start saying "but they're not caused by tides", then you deserve the response "nobody said they were, fool". If you stop deliberately misinterpretting things, you'll have a lot less stressful life. You probably think that my weight should only be measured in Newtons rather than kilograms, don't you? If so, yet again, you'd be wrong.
Maybe only 17 queries are required. So even if they did to some kind of rate-limitting to prevent mass sucks of account names, they'd not stop the leak.
Number all the names you're interested in binary. If you have 75000 names, then the binary numbers will be 17 bits long. In the first query, do a lookup on the (75000-65536) contacts which have a set 16th bit. Store all the results. In the second query, do a lookup on all the 32768 contacts which have a set 15th bit, again, store those. In the third query, do a lookup on all the (16384+16384) contacts which have a set 14th bit, again store. After 17 queries, each contact will be returned in exactly the sets which correspond to the bits that are set in its binary number, but not the others. I.e. it will be uniquely identifiable.
Of course, the fix for the problem is for the doofera at snapchat to simply not return account names in the query, and this 4000x speedup will stop working as quickly as the original. However, anyone who's done a huge suck prior to that could leak it out, so it must be considered that your account name is known to everyone. Expect more targetted adverspamming...
We are told (QI?) that US public phone booths do not have doors because a door would create a reasonable expectation of privacy. The implication is that if since that decision was made, you make a phone call from such a booth, all information about it is fair game to whomever. So your use of telecoms facilities has never been particularly private. (Personally, from that superficial summary above I think even Smith v. Maryland has gone too far. Why was there no warrant? Would that have required too much police-work? Like the "hunch"es that we see on the movies?)
I think the "how on earth can SCOTUS get away with that?!?!?" moment was when an ex-Monsanto lawyer thereon didn't declare an interest and then judged in favour of Monsanto.
Sure, they've been way more wrong (cough, Dred Scott, cough), but at least that was a clear and open wrongness (in that case, some of the most racist frothing I've ever read someone in a state of power utter), rather than a corrupt ulterior one.
> outside the US state means country
Just outside the US, to the south, you'll find the United Mexican States.
Time for a classic:
The United Nations initiated a poll with the request, "Please tell us your honest opinion about the lack of food in the rest of the world." The poll was a total failure. The Russians did not understand "Please". The Italians did not know the word "honest". The Chinese did not know what an "opinion" was. The Europeans did not know "lack", while the Africans did not know "food". Finally, the Americans didn't know anything about the "rest of the world".
Even "owning" might not be enough.
Do you not have housing associations? Where you "own", but you own a share in the association which grants you the right to live in the property that your share corresponds to.
Likewise, do you not have leasehold ownership, rather than freehold ownership? In the UK, almost everything that's "owned" is leasehold. (You own the building and the right to live there, but not the land its on.)
Do you also not have enough soft recycled wood pulp to use as toilet paper? Otherwise, why are all three branches of government using the constitution for that task?
And if you throw 10000 notes at the problem, you can take a photo where both the eye itself, which takes up a large proportion of the frame it is so close and the reflected scene about 1-2m away are in perfect focus, and with vivid colours to boot. One of the best photos I've seen in the last few years, and when my mate says "no photoshop apart from when gamut-mapping from raw to jpeg", I believe him. I'm still not sure I know exactly how he did it so brilliantly, I'm guessing it was very careful lighting.
Cheese? That's nearly a solved problem in AI, here's some recent state-of-the-art work:
http://www.youtube.com/watch?v=GzzOw2tmb3A
Sounds like you know more about what actually happens in the US than either me or target:
"""
Target does not have access to nor does it store the encryption key within our system. The PIN information is encrypted within Targetâ€(TM)s systems and can only be decrypted when it is received by our external, independent payment processor.
"""
-- http://pressroom.target.com/news/target-data-security-media-update-4
That first "encryption key" should read "decryption key", surely? How can they perform encryption within their system without the encryption key being stored somewhere within their system?
The whole system makes no sense to me. The chip on the card should verify the PIN. It should never be transmitted out of the terminal, definitely not sent to any third party, and it should certainly never be stored anywhere. Once the chip has verified the PIN, that yes/no answer is all you need to store. At least that's how things work in backwards eastern Europe.
Is there any reason to believe that whatever is required to get from the super-secret encryption key to the session encryption key (a terminal identifier, timestamp, nonce?) has not also been leaked along with the encrypted PIN and other data? Is the distinction between an "encryption" and a "decryption" key meaningful (is this symmetric crypto, the above quote is so mangled it's hard to trust anything in it?) - as even in the absense of a decryption key, an encryption key would permit the creation of a dictionary (or equivalenly a 10^4/10^6 brute force) for finding which PIN encrypted to the particular cyphertext.
> It allows for a reference point.
Now reread the paragraph you're responding to. It's hard for you to be more wrong. It's not a reference point, *meaningless*. People who work with databases will never have *any* interest in this toy. Ever. As someone currently looking at petabyte databases, I even laugh at the few-terabyte setups now. A tweak I was looking at this month managed to save a few terabytes in storage for one typical scenario. It's uncertain if that's significant enough to be worth implementing - more research required.
> tell me where on the ROI list partnering on academic research papers falls. It's not there
False. That's basically the entire raison d'etre for the creation of the last place where I worked, a supposed "R&D" centre. It's part of what my job was. So I'm speaking from experience of the cutting-edge IT field in 2013. They were happy to throw a good 7 figures in that direction. Feel free to ask them where the ROI was, all I know is that it happens because it happened.
> Have an interesting book buying list, travel [...]
Real world example I know of personally: Have a (nominally Christian) boyfriend from a country where the prevailing religion is Islam. Bang - straight onto the list. So much so that the pair in question even picked up a tail of spooks at least once when on holiday.
You're confusing "actually tides" and "like tides". The closest thing to a tsunami is a tide. Tsunamis are about as tidal as you can get without actually being tides.
Yes, you can model the uninterrupted propagation of tsunamis as waves. Yes, you can model the diffraction around island-sided opjects as waves. Yes, you can model the propagation of tsunamis down channels as waves, solitons as you say. But no, humans cannot realisitically use meta-material approaches to modifying tsunamis in order to protect against them.
Unless you can build *island-sized* features. The wavelenth is 500-1000km. The meta-material needs features with a size comparable to that of the wavelenth in order to affect wave-like behaviour in the ways being talked about. We do not have the capability to build such structures. I've not read the particular research the article refers to, but I presume it's to do with the negative refractive index lens. That permits you to create a shadow (i.e. shield what is behind) which is larger than the object itself. (So in theory, small things could protect an island.) However, in order to do that, you have to build a metamaterial structure the size of the shadow you wish to create. That's not a win. It's a massive loss, as now you have to build a huge fancy thing rather than a huge dumb thing.
They've also recruited at least 20 people whose response to "and what was your password?" is their password. It doesn't matter how many "brightest minds" you have if you have such weak links.
They did not publish a formal TPC-H result, they simply name-dropped TPC-H. In a way that, as it can't be compared to anything, TPC themselves is "meaningless".
And the 3 guys with nvidia email addresses on the paper have no interest in how well nvidia cards sell? It is you who seems confused, or at least naive. This was nvidia marketting. The company I just quit from used to do exactly the same - they all do.
Unless those processors are trying to use the same bus, in which case, more processors can mean more contention. Worst of all - by a superlinear amount.
Don't try to optimise until you've measured what you are most often doing with the box, in particular what you are most often waiting for. If you're very lucky, you might be compute bound, in which case, the more the merrier when it comes to CPUs (but only if the software is coded to scale sensibly). However, in my real-world experience (rather than hobby stuff), being I/O-bound is far more common.
Nothing new about that, according to a nuncle I once had, who used to wear a napron while eating a norange.
However, regarding Brits who write "should of", if I had my way, I'd have all of them shot.
"Hi! I'm the plumber who's going to fix the coolant pump"
<cue: cheey 70s synth music>
The greatest amount of damage from tsunamis happens when *other things* bend the tsunami. Anything which diverts the energy away from one thing diverts it towards another thing. Heaven help you if you've got two things diverting the energy towards you, such as sides of a bay, as then you're more likely to have an e^-kd rather than k.d^-2 reduction in energy (an extreme - compare losses in fibre optics to the inverse square law).
Thanks for the correction. I did check and realise my error after posting, but felt it unimportant, as the way the benchmarks had appeared, the H on was the one used on the at-the-time huger datasets. As time has progressed, they (TPC) now just say it's for "large" volumes of data, but they do clearly emphasise that the size is relevant. Maybe I picked up the folk meaning from a journo, and didn't bother verifying.
But anyway, even if they clustered 8 nodes together, which they didn't do, 96GB would still not cut it. The whole point of TPC was for real-world scenarios. These guys are still playing with toys. This is not a breakthrough, this is merely an interesting line of investigation. It might lead to a breakthrough, certainly, but they should only publish results when they're actually playing in the same ball-park, not 3 orders of magnitude away.
Yeah, but the solution to the problem in high energy physics is to make sure the high energy things don't touch anything important.
So, in this case, all we need to do is to levitate the ocean.
And you do realise that the mass of these pulses that LHC is trying to bend is only a tenth of a millionth of a gram? Tsunamis weigh quite a bit more than that.
But jesting aside, the article does look mostly bullshit, as it's highly inappropriate to model tsunamis as waves and solutions as working upon waves. They're so low frequency they're effectively DC. That's why in the term "tidal wave" is inappropriate - the "tidal" part is fine - in fact, it's almost perfect - it's the "wave" part that's misleading.
There's a weapon that works against such behaviour - being patronising. It only works slowly, but it does work eventually. Don't snark here - just laugh "useless use of pipes", or "useless use of cat", or "useless use of $whichever_thing_he_used_which_was_useless" every time you see it. It's a hundred times easier and quicker to laugh than it is to type a dumb command line, so it's a very cheap remedy.
I had a very bright co-worker who had never encountered the '-w' switch in grep. It took precisely one respectfully-made snark to fix that behaviour. He's optimised a few of the tasks that I commonly do too. Those who care learn, those who don't care get laughed at.