But even so, we could use multiple cores to speed those legacy machines up, Java, for instance: runs interpretatively until you have executed a block a certain number of times then "JIT" compiles it for a speed-up. But on a multicore machine, you should start other cores compiling in parallel with the interpreter so that you get to the compiled speedup sooner, and it can start prospectively downloading all the jars you need instead of waiting until you try and execute them and stalling.
Even with traditional languages, multicore needs a different approach. If you have a block of code which is small enough to fit into L1 cache, or possibly L2, then with an extra core it is "free" - it adds nothing to execution time. Rather than stopping to garbage collect, a garbage collector in an otherwise unused core can be optimising your heap all the time - for "free". Of course, the design of such a garbage collector is tricky - but it is a one-off design that will speed up every program that runs on the system.
You are right that current languages are badly designed for multicore. But look at the way in which the ghastly x86 CPU architecture has seen off far more elegant Risc architectures: a turbo-charged pig can, history shows, defeat an unimproved thoroughbred.
Twenty five years ago, I got the development system for an early approach to parallelism, Occam. This is old enough that it loaded, slowly, off 8-inch floppies. But unlike all the other software at the time (it was running on RT-11), it used internal parallelism since it was written in itself. Which meant that it came up and started working while the floppies were still clanking away loading the rest of the program. We need to treat main memory today as that system treated disk - as slow-access backing store.
I don't see anything in the referenced article about using the public internet for transmission of military data. He is talking about the public internet as a target for offensive action by an enemy with the intention of damaging the country and forcing a surrender, or at least achieving military advantage, by economic means.
Most purchasing is now done electronically, one way or another, and such purchases travel over the Internet. If you jam the Internet, you can seriously bollocks up a production and distribution based on "Just In Time". If all the food stores in the country cannot order fresh supplies, they will run out, and civilians, rightly or wrongly, will panic and demand that the Government Do Something - which may be to give in to the demands of whoever is jamming the system.
There are similar problems with a lot of monitoring equipment, on thing such as the power grid, which report over the Internet. Systems which have problems will not be able to report back to the maintenance base that they need attention. Of course, well designed systems will shut down if not fixed, so you wouldn't get any disasters immediately. But over time, large chunks of infrastructure would crumble. The crews who used to patrol the networks regularly have been laid off because automation - aided by the Internet - does it so much more cheaply.
It would be they cyber equivalent of taking out all the bridges in the road system - as was done in the Gulf Wars. By destroying public, not military, infrastructure, you can achieve military ends. And the military, charged with defending us, would rather this did not happen.
The military have their own separate networks, about which they do not talk. All the well publicised breaches in military network security have been in public-facing systems like public web servers and mail gateways. Yes, these have sometimes been pathetically badly secured. But they are not mission critical system - or at least, not intended to be.
I have broadband connections in two places. With my 10Mbit headline Virgin cable service, I get 9.6Mbit+ which persists long enough for me to download a Linux ISO. With the 8Mbit headline ADSL I can get about 5.5 Mbit for the same purpose. I suspect some upstream blocking, because when this line first came active, I was getting 7.6Mbit, but I haven't seen that for a year or so.
So you can get reasonable connections in some places.
That is the difference between PR agencies, an admittedly biased source working for their client, and AP, a supposedly unbiased wire service passing on original news material to media companies to use as they see fit.
What end publishers may do is their choice, but AP as a wire service intends to pass on originals. If they didn't draw a pretty hard line on this, then you could get multi-generation changes as each user "improves" the picture slightly before passing it on. AP intends to provide "raw" news without opinions, and "original" photos without touchup. What their customers do, whether politically or aesthetically, with the information AP provides is the customers business.
I think the "before" picture has been passed through a relatively high compression *since* being used to create the after, thus producing the colour shift and the artefacts.
Well, you do bring into question to what "life" actually is. But I think that mere recording is not observation. It is conceivable that an automaton could be able enough to interpret the observations, and thus constitute "life".
However, I think the point is irrelevant, because a universe in which an automaton could exist probably qualifies as one in which life could exist, regardless. The problem with the uncountable number of non-life-holding is that their physical laws are such that matter cannot exist, or that gravity is such that everything squelched into black holes in the first nanoseconds of the universe, or photons are so massive that they outweigh electrons or... We are not talking a little bit strange, we are talking very strange.
He didn't, in this answer, say anything about distributing software. He hoped that "nobody would be employed in developing or promoting proprietary software" - unqualified, so any where, any time, for any purpose. He wants all software, of any sort whatsoever, to be free. Of course, I am sure he realises as well as I do that is an unattainable ideal - but that is what he wants.
As to the second: I agree that the developer might well win the lawsuit - but I wouldn't be sure. My particular example is the manufacturer of a bicycle tyre valve, which was put by an unscrupulous cheapskate manufacturer into a motorbike, for which the valve manufacturer had stated it was unsuitable. A bike crashed due to this, the motorcycle manufacture had no money, and the valve manufacturer, who was insured, was forced to pay the damages. This despite an explicit disclaimer that the valve was not suitable for the purpose which caused the accident.
What price "not accepting liability" in this case? With ambulance chasing layers, it could cost millions to defend the suit, even if you win.
IANAL, of course - which makes the prospect all the more frightening. I have seen companies lose millions defending against baseless suits. How much is IBM actually going to recoup of its costs against SCO?
Stallman says " I hope to see the day when nobody is employed in developing or promoting proprietary software."
Cloud cuckoo land so far as the embedded software community is concerned.
At the extreme, the military are never going to open-source, say, missile guidance systems. The chance that the software structure or some embedded fact (e.g. that the range storage can only reach 1000 km) will leak some militarily significant fact.
But at a lower level, imagine someone open sources an Anti-lock Breaking System, and the system is involved in an accident in which it might, or might not, in its original or modified form, have contributed to an accident. Yes, I know that most open-source licenses carry complete disclaimers. The cost of defending the, when people's children are in the morgue is too high for a company to consider. US tort law, allowing claims against any party with money, is just too wide ranging.
No, the principle is still a principle. The principle is that "It is unsurprising that an observer finds his universe suitable for life since only universes suitable for life can contain observers". This is true regardless of how many universes exist and how many of them have life in them. The principle classifies meta-universal models into two classes: one in which there are many different universes, and the existence of life in some of them is unsurprising (but some cause for many universes must be given), and one in which there are few (e.g. one) universes, in which the existence of life is perhaps surprising and needs an explanation.
Essentially, this article is saying that current physical theory (with all the caveats about string theory being totally unproven) is pointing towards the former, in which no creator is needed, rather than the latter, in which case a creator is one hypothesis to explain why the univers is suitable for life as we know it.
The trouble with "saving students a ton of money" is that it would cost the publishers a ton of money. OK, half that money goes to the bookseller who, as a no-longer-needed intermediary would disappear. But the remainder is lost income for the publisher. The publisher will say that this will produce a dramatic drop in the number of textbooks they publish, to the consequent loss of the whole of academia. Whether that is true or not, if someone's income from a source drops dramatically, they are going to do less of it.
And if the University us paying a block fee for the e-textbook, they are going to have to get that money from somewhere. So either they will charge students for it or they will drop something, presumably worthwhile, that they are doing now (tuition, library books, formal dinners...).
The trouble with using correlations you don't understand is that they may be pure coincidence, in which case they will stop working "instantly". If you measure enough things, some of them are going to coincide for quite a while. Or the correlation may derive from some underlying mechanism which stops working for reasons you don't understand.
I was always worried by the "traditional" gene as one block of DNA coding for one protein. There simply wasn't enough code there to generate the complexity we see. When they say each such block cosed for an average of 5.7 proteins, plus lots of other info in methylation and teleomeres. Which means that there is much more information being transferred. Much more messily, but who said it had to be simple?
Has anyone done a cost/benefit analysis on this? Not only the cost of the original "black boxes" and their databases, but the cost of doing any effective analysis on the database, and the cost of following up any false positives you get from that analysis.
Remember that that the traffic they are looking for (allegedly), the real terrorist traffic, is a tiny, tiny part of the total net traffic. There are few terrorists, and even those that there are probably use the net legitimately as well as for terrorist purposes. So you hare going to have to dig deep to find them, and even a 99.99% accurate detector (unbelievably accurate) is going to turn up thousands of false positives for each real positive. The investigation time wasted will be mind boggling.
And that is without going into the civil liberties point, because the rest of/. will do it for me.
Praying to God for something physically impossible, and then for it to happen. If the sun stopped in the sky if, and only if, the Pope and 100,000 people prayed for it, that would be pretty convincing. A God who existed, and intervened regularly, would be easy to prove. It is the non-existence of God that cannot be prove: there can always be some chink of the Universe where you have not looked.
I recently attended a lecture by Ben Goldacre, author of the Bad Science column in the Guardian and book of the same name. He regularly debunks newspaper "experts", usually in the medical/health care/nutrition area. He gave numerous examples where the newspaper's so-called experts were, as I would see it, nothing of the sort. Without commenting on the particular case, most newspaper editors are scientific illiterates who will grace with "expert" anybody who knows anything at all about the subject.
In the UK, they merged the Revenue service, basically equivalent to the IRS, with the Social Services payout system, which pays money to the poor. Sounds like a good ides, because both parties need to know about your income do decide what tax you need to pay/subsidy you are awarded.
But actually it turned out a bit of a disaster because the Revenue is used to people with regular jobs and paychecks, and sorting out the balance at the end of the year. And the sort of people to whom a few hundred pounds under/over payment is not a disaster. But the Social Services clients are the sort of people who are in and out of jobs, and live from week to week. People who lost jobs didn't get their "safety-net" payments because the Revenue couldn't act fast enough, and ended up in serious trouble. And some people got over-paid, spent the money as they received it, and were threatened with starvation when the Revenue tried to claw back the overpayment.
The "efficiencies" didn't appear, and whole load of problems due to cultural differnces between different groups sprang up.
That depends upon the ability to get the foreclosure proceedings dismissed. I am surprised that is, in the long term, possible. In the UK, it can be fought for a while, and the courts will pressure the lender to do a deal, but foreclosure cannot be dismissed.
Even so, this does not affect my point either way. If what you describe is possible, it is possible whether the mortgage is held by the original lender of has been sold on. But selling on reduces the options for a negotiated climb down.
The "Jingle Mail" option only works in the US, which has "limited recourse". In the UK, the borrower is personally liable for any difference between the loan and what the lender gets from selling the foreclosed property. Which can be very painful, but removes the temptation to trash on exit.
Another factor which contributed was the Basle-1 and -2 capital adequacy requirements - particularly the latter. This is another example of the Law of Unintended Consequences. These rules were intended to force the assorted banks to have enough capital, in order to forestall exactly the liquidity crisis which has happened. But Basle-2 particularly allowed the riskiness of loans to be weighted so that less risky loans required less capital. So when the front line mortgage lenders reached their Basle limits, they invented the Securitised Investment Vehicles with weighted riskiness, and sold them on, thus recapitalising themselves for more lending. But the SIVs were badly designed and didn't pass the risk where everybody thought they did, leading to the melt-down. So a standard pit in to prevent the liquidity crisis actually contributed to it.
Another bad side effect of SIVs is making it much more difficult to be lenient on mortgages. When a lender forecloses, they typically only get about 50% of the original value of the home. Often the borrower trashes the house before they leave (a factor of US "limited recourse" mortgages"), and they they are selling in a saturated market. Much better, therefore, to write off, say, 25% of the mortgage without foreclosing and leave the borrower in possession. But if you have sold off the mortgage in many slices to several other investors, this becomes effectively impossible.
IMO, Google's behaviour in China is the least bad of the available options. To select it, therefore, is not "doing evil" even if the same actions in different circumstances would be evil. Cutting people with knives is usually evil - but not if you are doing life-improving surgery. If Google had the option of not censoring, then to censor would be evil. But it does not have that option; I can see no way it could open up that option; and therefore it is not evil.
The 90%, or a large slice of it, is nuclear and large coal fired power stations that are hard to turn on and off. These are the baseload stations, and they run 24/7. Then there are lighter-weight stations that can be turned on and off in an hour or two, which run during the day. Then you have some very lightweight stations using technologies such as gas turbines, which can spin up in seconds. These are turned on just at the peaks, and constitute the t10% which is rarely used.
Such systems cost money. It is probably more economical for the utility to install such systems, if they are worthwhile at all, than thousands of individual users. There are bound to be economies of scale.
See other replies: in the UK, liberalism has nothing to do with gun control. Any form of control is, in my mind, fundamentally illiberal - it removes liberty. It may be justified, but it is illiberal. Which is why we are confused by the US usage of the term to describe a political attitude which has very little to do with liberty and a lot to do with control.
My point, are the beginning of my post, was that is the recent, US-only, meaning of liberal. The older meaning, used by the rest of the English speaking world, says nothing about SUVs. It may say something about women's rights (they should have the same liberty as men, in all senses) and racism (ditto). A UK liberal says you are free to have your SUV provided you pay the true cost of owning it - which might put up the cost of gas to compensate for the pollution you create.
In the UK, "spreading the wealth around" is socialism, not liberalism. The two are different.
Slashdot Mirror
But even so, we could use multiple cores to speed those legacy machines up, Java, for instance: runs interpretatively until you have executed a block a certain number of times then "JIT" compiles it for a speed-up. But on a multicore machine, you should start other cores compiling in parallel with the interpreter so that you get to the compiled speedup sooner, and it can start prospectively downloading all the jars you need instead of waiting until you try and execute them and stalling.
Even with traditional languages, multicore needs a different approach. If you have a block of code which is small enough to fit into L1 cache, or possibly L2, then with an extra core it is "free" - it adds nothing to execution time. Rather than stopping to garbage collect, a garbage collector in an otherwise unused core can be optimising your heap all the time - for "free". Of course, the design of such a garbage collector is tricky - but it is a one-off design that will speed up every program that runs on the system.
You are right that current languages are badly designed for multicore. But look at the way in which the ghastly x86 CPU architecture has seen off far more elegant Risc architectures: a turbo-charged pig can, history shows, defeat an unimproved thoroughbred.
Twenty five years ago, I got the development system for an early approach to parallelism, Occam. This is old enough that it loaded, slowly, off 8-inch floppies. But unlike all the other software at the time (it was running on RT-11), it used internal parallelism since it was written in itself. Which meant that it came up and started working while the floppies were still clanking away loading the rest of the program. We need to treat main memory today as that system treated disk - as slow-access backing store.
I don't see anything in the referenced article about using the public internet for transmission of military data. He is talking about the public internet as a target for offensive action by an enemy with the intention of damaging the country and forcing a surrender, or at least achieving military advantage, by economic means.
Most purchasing is now done electronically, one way or another, and such purchases travel over the Internet. If you jam the Internet, you can seriously bollocks up a production and distribution based on "Just In Time". If all the food stores in the country cannot order fresh supplies, they will run out, and civilians, rightly or wrongly, will panic and demand that the Government Do Something - which may be to give in to the demands of whoever is jamming the system.
There are similar problems with a lot of monitoring equipment, on thing such as the power grid, which report over the Internet. Systems which have problems will not be able to report back to the maintenance base that they need attention. Of course, well designed systems will shut down if not fixed, so you wouldn't get any disasters immediately. But over time, large chunks of infrastructure would crumble. The crews who used to patrol the networks regularly have been laid off because automation - aided by the Internet - does it so much more cheaply.
It would be they cyber equivalent of taking out all the bridges in the road system - as was done in the Gulf Wars. By destroying public, not military, infrastructure, you can achieve military ends. And the military, charged with defending us, would rather this did not happen.
The military have their own separate networks, about which they do not talk. All the well publicised breaches in military network security have been in public-facing systems like public web servers and mail gateways. Yes, these have sometimes been pathetically badly secured. But they are not mission critical system - or at least, not intended to be.
I have broadband connections in two places. With my 10Mbit headline Virgin cable service, I get 9.6Mbit+ which persists long enough for me to download a Linux ISO. With the 8Mbit headline ADSL I can get about 5.5 Mbit for the same purpose. I suspect some upstream blocking, because when this line first came active, I was getting 7.6Mbit, but I haven't seen that for a year or so.
So you can get reasonable connections in some places.
The tool that finally got the jammed Skylab solar "wings" to unfold was a $10 crowbar from the local hardware store in Florida.
That is the difference between PR agencies, an admittedly biased source working for their client, and AP, a supposedly unbiased wire service passing on original news material to media companies to use as they see fit.
What end publishers may do is their choice, but AP as a wire service intends to pass on originals. If they didn't draw a pretty hard line on this, then you could get multi-generation changes as each user "improves" the picture slightly before passing it on. AP intends to provide "raw" news without opinions, and "original" photos without touchup. What their customers do, whether politically or aesthetically, with the information AP provides is the customers business.
I think the "before" picture has been passed through a relatively high compression *since* being used to create the after, thus producing the colour shift and the artefacts.
Well, you do bring into question to what "life" actually is. But I think that mere recording is not observation. It is conceivable that an automaton could be able enough to interpret the observations, and thus constitute "life".
However, I think the point is irrelevant, because a universe in which an automaton could exist probably qualifies as one in which life could exist, regardless. The problem with the uncountable number of non-life-holding is that their physical laws are such that matter cannot exist, or that gravity is such that everything squelched into black holes in the first nanoseconds of the universe, or photons are so massive that they outweigh electrons or... We are not talking a little bit strange, we are talking very strange.
He didn't, in this answer, say anything about distributing software. He hoped that "nobody would be employed in developing or promoting proprietary software" - unqualified, so any where, any time, for any purpose. He wants all software, of any sort whatsoever, to be free. Of course, I am sure he realises as well as I do that is an unattainable ideal - but that is what he wants.
As to the second: I agree that the developer might well win the lawsuit - but I wouldn't be sure. My particular example is the manufacturer of a bicycle tyre valve, which was put by an unscrupulous cheapskate manufacturer into a motorbike, for which the valve manufacturer had stated it was unsuitable. A bike crashed due to this, the motorcycle manufacture had no money, and the valve manufacturer, who was insured, was forced to pay the damages. This despite an explicit disclaimer that the valve was not suitable for the purpose which caused the accident.
What price "not accepting liability" in this case? With ambulance chasing layers, it could cost millions to defend the suit, even if you win.
IANAL, of course - which makes the prospect all the more frightening. I have seen companies lose millions defending against baseless suits. How much is IBM actually going to recoup of its costs against SCO?
Stallman says " I hope to see the day when nobody is employed in developing or promoting proprietary software."
Cloud cuckoo land so far as the embedded software community is concerned.
At the extreme, the military are never going to open-source, say, missile guidance systems. The chance that the software structure or some embedded fact (e.g. that the range storage can only reach 1000 km) will leak some militarily significant fact.
But at a lower level, imagine someone open sources an Anti-lock Breaking System, and the system is involved in an accident in which it might, or might not, in its original or modified form, have contributed to an accident. Yes, I know that most open-source licenses carry complete disclaimers. The cost of defending the, when people's children are in the morgue is too high for a company to consider. US tort law, allowing claims against any party with money, is just too wide ranging.
No, the principle is still a principle. The principle is that "It is unsurprising that an observer finds his universe suitable for life since only universes suitable for life can contain observers". This is true regardless of how many universes exist and how many of them have life in them. The principle classifies meta-universal models into two classes: one in which there are many different universes, and the existence of life in some of them is unsurprising (but some cause for many universes must be given), and one in which there are few (e.g. one) universes, in which the existence of life is perhaps surprising and needs an explanation.
Essentially, this article is saying that current physical theory (with all the caveats about string theory being totally unproven) is pointing towards the former, in which no creator is needed, rather than the latter, in which case a creator is one hypothesis to explain why the univers is suitable for life as we know it.
The trouble with "saving students a ton of money" is that it would cost the publishers a ton of money. OK, half that money goes to the bookseller who, as a no-longer-needed intermediary would disappear. But the remainder is lost income for the publisher. The publisher will say that this will produce a dramatic drop in the number of textbooks they publish, to the consequent loss of the whole of academia. Whether that is true or not, if someone's income from a source drops dramatically, they are going to do less of it.
And if the University us paying a block fee for the e-textbook, they are going to have to get that money from somewhere. So either they will charge students for it or they will drop something, presumably worthwhile, that they are doing now (tuition, library books, formal dinners...).
The trouble with using correlations you don't understand is that they may be pure coincidence, in which case they will stop working "instantly". If you measure enough things, some of them are going to coincide for quite a while. Or the correlation may derive from some underlying mechanism which stops working for reasons you don't understand.
I was always worried by the "traditional" gene as one block of DNA coding for one protein. There simply wasn't enough code there to generate the complexity we see. When they say each such block cosed for an average of 5.7 proteins, plus lots of other info in methylation and teleomeres. Which means that there is much more information being transferred. Much more messily, but who said it had to be simple?
Absolutely agreed.
Has anyone done a cost/benefit analysis on this? Not only the cost of the original "black boxes" and their databases, but the cost of doing any effective analysis on the database, and the cost of following up any false positives you get from that analysis.
Remember that that the traffic they are looking for (allegedly), the real terrorist traffic, is a tiny, tiny part of the total net traffic. There are few terrorists, and even those that there are probably use the net legitimately as well as for terrorist purposes. So you hare going to have to dig deep to find them, and even a 99.99% accurate detector (unbelievably accurate) is going to turn up thousands of false positives for each real positive. The investigation time wasted will be mind boggling.
And that is without going into the civil liberties point, because the rest of /. will do it for me.
Praying to God for something physically impossible, and then for it to happen. If the sun stopped in the sky if, and only if, the Pope and 100,000 people prayed for it, that would be pretty convincing. A God who existed, and intervened regularly, would be easy to prove. It is the non-existence of God that cannot be prove: there can always be some chink of the Universe where you have not looked.
I recently attended a lecture by Ben Goldacre, author of the Bad Science column in the Guardian and book of the same name. He regularly debunks newspaper "experts", usually in the medical/health care/nutrition area. He gave numerous examples where the newspaper's so-called experts were, as I would see it, nothing of the sort. Without commenting on the particular case, most newspaper editors are scientific illiterates who will grace with "expert" anybody who knows anything at all about the subject.
And if you enter a cloud, you usually get lost. Flying in clouds is regarded as very bad thing if you can void it.
In the UK, they merged the Revenue service, basically equivalent to the IRS, with the Social Services payout system, which pays money to the poor. Sounds like a good ides, because both parties need to know about your income do decide what tax you need to pay/subsidy you are awarded.
But actually it turned out a bit of a disaster because the Revenue is used to people with regular jobs and paychecks, and sorting out the balance at the end of the year. And the sort of people to whom a few hundred pounds under/over payment is not a disaster. But the Social Services clients are the sort of people who are in and out of jobs, and live from week to week. People who lost jobs didn't get their "safety-net" payments because the Revenue couldn't act fast enough, and ended up in serious trouble. And some people got over-paid, spent the money as they received it, and were threatened with starvation when the Revenue tried to claw back the overpayment.
The "efficiencies" didn't appear, and whole load of problems due to cultural differnces between different groups sprang up.
That depends upon the ability to get the foreclosure proceedings dismissed. I am surprised that is, in the long term, possible. In the UK, it can be fought for a while, and the courts will pressure the lender to do a deal, but foreclosure cannot be dismissed.
Even so, this does not affect my point either way. If what you describe is possible, it is possible whether the mortgage is held by the original lender of has been sold on. But selling on reduces the options for a negotiated climb down.
The "Jingle Mail" option only works in the US, which has "limited recourse". In the UK, the borrower is personally liable for any difference between the loan and what the lender gets from selling the foreclosed property. Which can be very painful, but removes the temptation to trash on exit.
Another factor which contributed was the Basle-1 and -2 capital adequacy requirements - particularly the latter. This is another example of the Law of Unintended Consequences. These rules were intended to force the assorted banks to have enough capital, in order to forestall exactly the liquidity crisis which has happened. But Basle-2 particularly allowed the riskiness of loans to be weighted so that less risky loans required less capital. So when the front line mortgage lenders reached their Basle limits, they invented the Securitised Investment Vehicles with weighted riskiness, and sold them on, thus recapitalising themselves for more lending. But the SIVs were badly designed and didn't pass the risk where everybody thought they did, leading to the melt-down. So a standard pit in to prevent the liquidity crisis actually contributed to it.
Another bad side effect of SIVs is making it much more difficult to be lenient on mortgages. When a lender forecloses, they typically only get about 50% of the original value of the home. Often the borrower trashes the house before they leave (a factor of US "limited recourse" mortgages"), and they they are selling in a saturated market. Much better, therefore, to write off, say, 25% of the mortgage without foreclosing and leave the borrower in possession. But if you have sold off the mortgage in many slices to several other investors, this becomes effectively impossible.
IMO, Google's behaviour in China is the least bad of the available options. To select it, therefore, is not "doing evil" even if the same actions in different circumstances would be evil. Cutting people with knives is usually evil - but not if you are doing life-improving surgery. If Google had the option of not censoring, then to censor would be evil. But it does not have that option; I can see no way it could open up that option; and therefore it is not evil.
The 90%, or a large slice of it, is nuclear and large coal fired power stations that are hard to turn on and off. These are the baseload stations, and they run 24/7. Then there are lighter-weight stations that can be turned on and off in an hour or two, which run during the day. Then you have some very lightweight stations using technologies such as gas turbines, which can spin up in seconds. These are turned on just at the peaks, and constitute the t10% which is rarely used.
Such systems cost money. It is probably more economical for the utility to install such systems, if they are worthwhile at all, than thousands of individual users. There are bound to be economies of scale.
See other replies: in the UK, liberalism has nothing to do with gun control. Any form of control is, in my mind, fundamentally illiberal - it removes liberty. It may be justified, but it is illiberal. Which is why we are confused by the US usage of the term to describe a political attitude which has very little to do with liberty and a lot to do with control.
My point, are the beginning of my post, was that is the recent, US-only, meaning of liberal. The older meaning, used by the rest of the English speaking world, says nothing about SUVs. It may say something about women's rights (they should have the same liberty as men, in all senses) and racism (ditto). A UK liberal says you are free to have your SUV provided you pay the true cost of owning it - which might put up the cost of gas to compensate for the pollution you create.
In the UK, "spreading the wealth around" is socialism, not liberalism. The two are different.