The GSM radio wavelength is about 30cm which means that in effect all objects which affect the radio path, including the transmitter and LED receiver, are "blurry" in space to the scale of 30cm (this is an order of magnitude, not an exact value). The phone itself, and the distance from the LEDs, are much smaller than that. So the directionality of the radiation is nearly irrelevant to calculating how much is absorbed and transmitted.
In other words, contrary to the parent post, the LEDs attached to the phone will be effectively on the radio path to the base station, no matter where they are attached on the phone.
It's counterintuitive that you can have a radio signal between two small antennae at A and B, and something that's nearly in between but off by say 10cm affecting the signal between A and B, is though attracting the energy towards it (even bending the beam is possible). But that is exactly what happens. Waves are like that.
It's more complicated than that, however, because the LEDs are also in the "near field" - the region where there may be a non-radiating component to the oscillating EM field around the phone transmitter. In this region, the LEDs could, if they are constructed to do so, absorb energy from the near field, and, depending very much on the phone design, potentially do it without affecting the radiated signal.
Also, it is possible that they absorb some of the radiated energy but if they use very little power, not affect it very much.
So we can't easily say what effect the LEDs will have on the transmitted signal, but the parent's argument about having to be "exactly on the path" to the transmitter, as in a straight line, is not correct.
I got it from the instruction manual of a microwave oven 20 years ago, which said something very similar to this USDA page.
I have always taken that to mean the microwave-induced molecular excitations take a while to thermalise (i.e. to settle into a thermal distribution of energy states, which would affect the measured temperature), but looking around the web there are many places which say that "standing time" is merely to redistribute the heat as you say. Both explanations are plausible on the face of it.
In fact, microwave ovens don't just heat things up.
When you remove food from a microwave oven, it continues getting hotter for a little while after leaving the oven. That's not just heat spreading through the food - it really does get hotter. Think about what must be going on in the food for that to happen - it certainly isn't the same as blasting it with heat.
Maybe the radiation from broadcast TV and Radio is harmful after all - maybe it is causing the stupidity you mention....
For that matter, maybe the radiation from the sun causes stupidity too (why not, it causes skin cancer so I hear)...
More interesting to me is, how will we ever find about those effects, if they occur? Nobody's turning off the TV broadcasts and assessing change in behavioural intellegince over the following decades, are they?
I agree with you about the cellphones. If anything, they ought to have Wi-Fi cellphones that use the school's network when within range, to reduce their power levels.
Any interaction for which all of the molecules rotate in the same direction is not a thermal interaction
Why not - rotational energy *is* themal energy.
No, rotational energy is not thermal energy in general. For example, a spinning wheel may have a lot of rotational energy, but that rotation does not necessarily contribute at all to the temperature of the wheel - in other words, it's non-thermal rotational energy.
The same goes even for spinning molecules - to the extent the spins are correlated with other spins, or predictable in any other way.
The reason why is because 'thermal interaction' as the GP uses it specifically means interactions between molecules (or larger/smaller groupings, as you like) where the energies of each interacting component are independently subject to the appropriate thermodynamic probability distribution for the temperature of the system. That's when you can apply the laws of statistical mechanics, and when the concepts of heat and temperature actually work.
Heat is disordered energy. The distribution of thermal energy among components of the system obeys various laws of statistical mechanics. E.g. increasing kinetic, vibrational and rotational energy in molecules as a substance increases with temperature.
Ordered energy is not called heat. For example, no matter how fast a block of ice moves, even though kinetic energy increases with speed, we don't say that it's temperature increases with speed.
That's because the increasing kinetic energy with speed in that case is highly ordered - every molecule gains speed equally in the same direction (although in general, any kind of order applies).
The scale doesn't have much to do with this. Spinning molecules can be ordered or not, and that makes a difference as to whether the rotational energy is thermal or not (or more realistically, how much of it you would describe as thermal).
To some extent, the concept of thermal does depend on your perspective. A million rotating wheels, for example, may start out ordered but after bouncing off each other for a while, would soon acquire a thermal energy distribution, and you can apply thermodynamics then. (I've seen this happen in video game simulations:)
By the way, everything I just said is not at all formally rigorous - because you can't do thermodynamics while reasoning about individual molecules. But it's basically what is meant be thermal vs. non-thermal energy.
If people have psychosomatic symptoms due to these wireless signals, those can still be harmful and debilitating, and in the long term some symptoms can induce detectable physical damage.
And often, you can't treat the symptom by simply declaring it as 'imaginary pain'. That (usually) doesn't help what is, of course, a real problem.
On a totally different note, how about this as an idea (it's just an idea, I don't believe it myself): what if the wireless signals are sensed by the body, causing no physical harm but getting sensed as an irritating signal (like some sounds, even when quiet)... and manifesting as the subjective experience of pain, fatigue, etc... Would that be 'real' or 'imaginary'? What is the appropriate response?:)
Large enough for the GP to reasonably adjust his/her computer usage for his/her own comfort (and possibly health, but that is harder to assess).
Not large enough to generalise to other people. But large enough to offer a suggestion that others with similar symptoms might try for themselves, and decide for themselves if it helps them.
Not large or controlled enough to tell whether the effect is due to radio emissions.
(My guess is the heat and/or vibration from the laptop caused the thigh pain and headaches. But it's just a guess.)
I dream for at least a third of the day while awake, in addition to sleep time (which is currently about 8 hours, but varies from 2 hours to 20 hours per 24 hours at extremes; yes I have problem sleep).
Not sleeping (or sleeping much less) wouldn't mean I stop dreaming, but it might mean I got more things done that I'd like to do in between dreams.
That said, day dreams are of quite a different quality, being more consciously led than night dreams.
I've watched people enter Chinese texts on cellphones in China, and it's amazing to watch. Whatever entry method they're using seems remarkably efficient.
I was using RBLs, but I'm actually getting better results since I disabled them. They were flagging too many legitimate mail senders, and they weren't very accurate anyway.
I think that's because several of them went out of date and stopped being trustworthy around the same time.
Another thing to have to keep updating/monitoring, and having to re-estimate the quality of responses from time to time. I hate spam.
Also, it's not acceptable to regularly bounce mails where a business correspondent is asked to send from a different site, especially if resending from the same place (*their* own domain) does not work. It looks unprofessional, and they're likely to phone and be unimpressed if it's important, or go elsewhere if someone else can do the same work without the hassle. Neither is helpful to me.
Oh I agree, Intel's support is much better in principle.
However, I've seen so many people talk about Intel's great video support, and Intel's own press releases seem to imply it, that I felt I had to share my experience: it's far from great at the moment. It's not actually Intel's fault; the main problem seems to be limited resource for X driver development (which has good 3D support when it works, but really dubious hacks elsewhere), and 3rd party hardware manufacturers who make broken video BIOSes.
For me and many others, even basic non-accelerated rendering doesn't work out of the box. For example, I'd heard great things about Ubuntu and it was very disappointing that the installation process led to a black screen before it was finished, and after a bit of typing blind and a reboot, resulted in a working display but non-native resolution on the laptop.
From all accounts, ATI and Nvidia's drivers, although shamefully binary, do work quite well for most people. I'd guess their drivers are also full of dubious hacks, but people say they seem to work quite well.
This would be an example of nice in principle, but disappointing implementation so far. And, although in principle "much easier" to get it all working, in practice it could take many years, if the experience of my previous laptop is anything to go by.
Intel are to be applauded for their open source move. But that's not enough by itself to make good, or even usable-out-of-the-box drivers. I really hope they can follow through with actually fixing the code...
(Btw, it has nothing to do with different variants of Linux. X drivers are binary-compatible even between different OSes, for the same Xorg version and machine architecture. Motherboard hardware and video BIOS variations, that's where the compatibility trouble lies.)
Nothing short of human tests will ever put an end to Spam.
Unfortunately, even human tests do not work. I occasionally discard a legitimate mail by accident - a false positive if you like - while I'm manually scanning through the 1000 or so daily messages in my inbox. (Another 5000 is automatically discarded by filters). This is because it takes such a long time to get through 1000 that each one gets only a moment's attention.
Also, there was a study about a year ago that concluded that some kinds of machine filtering have a lower false positive rate than humans.
I get about 6k a day to my aging server, and the spam filtering cannot keep up. It's not an even trickle: some times of the day, it's several attempts per second. That's faster than the filtering software can handle.
Legitimate senders are getting "warning: could not send for past 4 hours" and then phoning me to ask if I've received their mail. The CPU and memory load spikes from time to time, and then it's not possible to login until it settles. (A known weak spot in Linux.) If I lower the resources allocated to mail processing, it cannot handle the incoming mail rate most of the time - it's just on the edge right now.
Now, I'm running Sendmail and Spamassassin on a Red Hat 9 box which is a 600MHz Celeron in a data centre. I'm sure all of you will laugh, and tell me to run better software on a newer OS and a better PC with a virtual machine. And how SpamFilter-of-the-day is much better than Spamassassin (it could hardly be worse).
But the fact is, updating all of those takes real time and expense. And when it's updated? It'll still be a significant load, and it'll still need to be maintained, and upgraded again at some point. I'm planning to order a server with 2GB of RAM because I think 1GB won't be enough to handle the memory load spikes for spam filtering.
Sorting all that will cost me real time and money. But at least I'll have reliable mail again and a server I can use for other things.
And that's nothing compared with the time I spend every fucking day skipping the spams in my inbox. About 5k per day are deleted by my filters. But that leaves 1k per day to skip manually - those where the risk of false positive is high enough that I need to check them. That's a pretty long and unpleasant inbox to face each day. Very unpleasant if I don't check it for a few days.
However, that has got to the point where _I'm_ accidentally deleting legitimate mails too. I have taken to whitelist-scoring my most important correspondents so that I don't accidently delete them among the spams. But that makes it even harder to respond to mail from people that have no prior connection with me, and people who haven't made it to the whitelist but should have.
Spam is definitely a financial problem for me. I estimate it costs me about $15000 USD/year in time spent deleting unwanted messages (about 40 minutes/day). Add on the hardware and software maintenance costs, and the annoyance, and the problems caused by deleting false positives.
For some definition of "full support". Let's see... I have Intel 945GM video in my laptop, and the latest Xorg drivers. (I have tried several experimental driver versions too).
Q: Does 3d acceleration work? A: Only on the laptop's internal screen. There is no 3d acceleration when using an external DVI screen - turning on acceleration causes video initialization to fail with an obscure should-never-happen error.
Q: Does dual display (laptop and external screen together) work? A: Only for some resolutions (i.e. not the actual size of the external TFT), and only with some driver versions and rather picky Xorg.conf settings (i.e. extremely flaky), and only with all hardware acceleration disabled and software cursor, and even when it works it's a most unpleasant and sometimes buggy hack (initializes whole video card twice; two driver instances both poking the same registers without synchronisation). The version I'm using at the moment (Ubuntu Edgy's "intel" driver) doesn't work with dual display, though it is supposed to. I'd use the other Intel driver ("i810") but it doesn't work on this laptop with a DVI external screen.
Q: Does simply displaying at a screen's native resolution work?! A: For modern TFT resolutions (mine are 1280x800 and 1680x1050): only with an experimental driver, or a BIOS hack (it temporarily modifies tables in the BIOS). Neither of these works with a laptop's internal and an external display simultaneously - you have to pick one when X starts, switching to the other requires X to be restarted, all applications closed etc.
Q: Does suspend work? A: No, after suspend-to-RAM, the screen is blank. (I've not tried suspend-to-disk because I use encrypted swap, and suspend-to-disk doesn't work with that.)
But to its credit, with the appropriate hacks and experimental drivers, displaying on one screen only with acceleration disabled and software cursor seems to be good... Not sure I'd call that "full support" though! I'm looking forward to it when it arrives. Intel have got the right idea, it's only implementation which is lacking now, and that appears to be outside Intel's control, being more of an under-resourced Xorg thing.
The thing is, we have a highly distributed database, the network is not constantly available, and updates (within constraints) may happen simultaneously at different places in the network and be reconciled later. And most of the nodes run on small embedded systems, while some nodes are large, centralised severs.
So many of the concepts you suggest are probably quite relevant theory, but we can't work using the common database implementations.
I'm not so concerned with how it's "mapped to physical tables". I regard that, since we're not using standard database engines, as storage and indexing optimisation. Important, but it's not really modelling.
I would like to read more about the things you mention, especially data modelling as a discipline, and the things you listed: "dynamic attributes", "snowflake schema" etc. Suggest any other good books?
To be honest I'm working on a project where the data model is rather messy, and I designed it. But there is no _obvious_ right way to model it which fits with the ways we need to be able to access the data (there are real-time and distributed and offline aspects which mean we can't just use an RDB and standard normalisations).
Figuring out a good way forward would involve a couple of months break in visible features... and right now (as of a while), it's the features that are in constant demand. We're effectively writing a number of applications for different customers simultaneously, with most of the code in common, while we figure out what to do with insufficient resources to do the job 'properly'.
I suspect it's a common scenario.
And to be honest, we are servicing the business's short term needs quite well.
But I have to admit, the data model has plenty of room for improvement.
*rushes off to mock up a customer-specific feature...*
Will they start putting different content on each layer?
Extra scenes, only for licensed Blu-Ray users?
I can see this happening with data disks, to double the storage density if they're cheap enough to produce, and if they aren't superceded soon by another, higher-density format.
On the contrary, modern CPUs have had RET-prediction for about 10 years now, and RET-prediction is almost always correct, so there's little or no pipeline penalty. It's better to have subroutines if they save cache space, than to flatten out code. That's why you see some routines being uninlined in the Linux kernel source, these days. There is still the overhead of setting up call arguments etc., so it's not always gainful. Register arguments help with that.
A practical difference between MIPS and ARM is that there are a number of MIPS clones which don't have to license anything from MIPS, provided they leave out the patented instructions. Whereas, cloning an ARM gets lots of hassle from ARM Ltd. Even just writing a software simulator for ARM is a problem(!) - this is why Qemu only emulates old ARMs.
So if someone's going to implement one of those instruction sets, it will tend to be MIPS if they're designing their own chip, and either of them if they're buying a pre-designed core.
No, often you can find an instruction to put after the branch. Sparc is the same. It's called the "branch delay slot", and it's a way of reducing the pipeline flush penalty.
Slashdot Mirror
No, radio does not behave like that.
The GSM radio wavelength is about 30cm which means that in effect all objects which affect the radio path, including the transmitter and LED receiver, are "blurry" in space to the scale of 30cm (this is an order of magnitude, not an exact value). The phone itself, and the distance from the LEDs, are much smaller than that. So the directionality of the radiation is nearly irrelevant to calculating how much is absorbed and transmitted.
In other words, contrary to the parent post, the LEDs attached to the phone will be effectively on the radio path to the base station, no matter where they are attached on the phone.
It's counterintuitive that you can have a radio signal between two small antennae at A and B, and something that's nearly in between but off by say 10cm affecting the signal between A and B, is though attracting the energy towards it (even bending the beam is possible). But that is exactly what happens. Waves are like that.
It's more complicated than that, however, because the LEDs are also in the "near field" - the region where there may be a non-radiating component to the oscillating EM field around the phone transmitter. In this region, the LEDs could, if they are constructed to do so, absorb energy from the near field, and, depending very much on the phone design, potentially do it without affecting the radiated signal.
Also, it is possible that they absorb some of the radiated energy but if they use very little power, not affect it very much.
So we can't easily say what effect the LEDs will have on the transmitted signal, but the parent's argument about having to be "exactly on the path" to the transmitter, as in a straight line, is not correct.
-- Jamie
I got it from the instruction manual of a microwave oven 20 years ago, which said something very similar to this USDA page.
I have always taken that to mean the microwave-induced molecular excitations take a while to thermalise (i.e. to settle into a thermal distribution of energy states, which would affect the measured temperature), but looking around the web there are many places which say that "standing time" is merely to redistribute the heat as you say. Both explanations are plausible on the face of it.
-- Jamie
An obvious extension of your idea is to competitive sex games, with wiimotes tracking your pelvis thrusts to measure those elusive parameters.
-- Jamie
They also have materials which reflect radio, so depending on the local configuration, in places it can drop slower than inverse square...
-- Jamie
In fact, microwave ovens don't just heat things up.
When you remove food from a microwave oven, it continues getting hotter for a little while after leaving the oven. That's not just heat spreading through the food - it really does get hotter. Think about what must be going on in the food for that to happen - it certainly isn't the same as blasting it with heat.
-- Jamie
Maybe the radiation from broadcast TV and Radio is harmful after all - maybe it is causing the stupidity you mention....
For that matter, maybe the radiation from the sun causes stupidity too (why not, it causes skin cancer so I hear)...
More interesting to me is, how will we ever find about those effects, if they occur? Nobody's turning off the TV broadcasts and assessing change in behavioural intellegince over the following decades, are they?
I agree with you about the cellphones. If anything, they ought to have Wi-Fi cellphones that use the school's network when within range, to reduce their power levels.
-- Jamie
No, rotational energy is not thermal energy in general. For example, a spinning wheel may have a lot of rotational energy, but that rotation does not necessarily contribute at all to the temperature of the wheel - in other words, it's non-thermal rotational energy.
The same goes even for spinning molecules - to the extent the spins are correlated with other spins, or predictable in any other way.
The reason why is because 'thermal interaction' as the GP uses it specifically means interactions between molecules (or larger/smaller groupings, as you like) where the energies of each interacting component are independently subject to the appropriate thermodynamic probability distribution for the temperature of the system. That's when you can apply the laws of statistical mechanics, and when the concepts of heat and temperature actually work.
Heat is disordered energy. The distribution of thermal energy among components of the system obeys various laws of statistical mechanics. E.g. increasing kinetic, vibrational and rotational energy in molecules as a substance increases with temperature.
Ordered energy is not called heat. For example, no matter how fast a block of ice moves, even though kinetic energy increases with speed, we don't say that it's temperature increases with speed.
That's because the increasing kinetic energy with speed in that case is highly ordered - every molecule gains speed equally in the same direction (although in general, any kind of order applies).
The scale doesn't have much to do with this. Spinning molecules can be ordered or not, and that makes a difference as to whether the rotational energy is thermal or not (or more realistically, how much of it you would describe as thermal).
To some extent, the concept of thermal does depend on your perspective. A million rotating wheels, for example, may start out ordered but after bouncing off each other for a while, would soon acquire a thermal energy distribution, and you can apply thermodynamics then. (I've seen this happen in video game simulations :)
By the way, everything I just said is not at all formally rigorous - because you can't do thermodynamics while reasoning about individual molecules. But it's basically what is meant be thermal vs. non-thermal energy.
-- JamieNot if the subjects consent, which they could in this case.
-- Jamie
If people have psychosomatic symptoms due to these wireless signals, those can still be harmful and debilitating, and in the long term some symptoms can induce detectable physical damage.
:)
And often, you can't treat the symptom by simply declaring it as 'imaginary pain'. That (usually) doesn't help what is, of course, a real problem.
On a totally different note, how about this as an idea (it's just an idea, I don't believe it myself): what if the wireless signals are sensed by the body, causing no physical harm but getting sensed as an irritating signal (like some sounds, even when quiet)... and manifesting as the subjective experience of pain, fatigue, etc... Would that be 'real' or 'imaginary'? What is the appropriate response?
-- Jamie
Large enough for the GP to reasonably adjust his/her computer usage for his/her own comfort (and possibly health, but that is harder to assess).
Not large enough to generalise to other people. But large enough to offer a suggestion that others with similar symptoms might try for themselves, and decide for themselves if it helps them.
Not large or controlled enough to tell whether the effect is due to radio emissions.
(My guess is the heat and/or vibration from the laptop caused the thigh pain and headaches. But it's just a guess.)
I dream for at least a third of the day while awake, in addition to sleep time (which is currently about 8 hours, but varies from 2 hours to 20 hours per 24 hours at extremes; yes I have problem sleep).
Not sleeping (or sleeping much less) wouldn't mean I stop dreaming, but it might mean I got more things done that I'd like to do in between dreams.
That said, day dreams are of quite a different quality, being more consciously led than night dreams.
I've watched people enter Chinese texts on cellphones in China, and it's amazing to watch. Whatever entry method they're using seems remarkably efficient.
I was using RBLs, but I'm actually getting better results since I disabled them. They were flagging too many legitimate mail senders, and they weren't very accurate anyway.
I think that's because several of them went out of date and stopped being trustworthy around the same time.
Another thing to have to keep updating/monitoring, and having to re-estimate the quality of responses from time to time. I hate spam.
Also, it's not acceptable to regularly bounce mails where a business correspondent is asked to send from a different site, especially if resending from the same place (*their* own domain) does not work. It looks unprofessional, and they're likely to phone and be unimpressed if it's important, or go elsewhere if someone else can do the same work without the hassle. Neither is helpful to me.
Oh I agree, Intel's support is much better in principle.
However, I've seen so many people talk about Intel's great video support, and Intel's own press releases seem to imply it, that I felt I had to share my experience: it's far from great at the moment. It's not actually Intel's fault; the main problem seems to be limited resource for X driver development (which has good 3D support when it works, but really dubious hacks elsewhere), and 3rd party hardware manufacturers who make broken video BIOSes.
For me and many others, even basic non-accelerated rendering doesn't work out of the box. For example, I'd heard great things about Ubuntu and it was very disappointing that the installation process led to a black screen before it was finished, and after a bit of typing blind and a reboot, resulted in a working display but non-native resolution on the laptop.
From all accounts, ATI and Nvidia's drivers, although shamefully binary, do work quite well for most people. I'd guess their drivers are also full of dubious hacks, but people say they seem to work quite well.
This would be an example of nice in principle, but disappointing implementation so far. And, although in principle "much easier" to get it all working, in practice it could take many years, if the experience of my previous laptop is anything to go by.
Intel are to be applauded for their open source move. But that's not enough by itself to make good, or even usable-out-of-the-box drivers. I really hope they can follow through with actually fixing the code...
(Btw, it has nothing to do with different variants of Linux. X drivers are binary-compatible even between different OSes, for the same Xorg version and machine architecture. Motherboard hardware and video BIOS variations, that's where the compatibility trouble lies.)
-- Jamie
Unfortunately, even human tests do not work. I occasionally discard a legitimate mail by accident - a false positive if you like - while I'm manually scanning through the 1000 or so daily messages in my inbox. (Another 5000 is automatically discarded by filters). This is because it takes such a long time to get through 1000 that each one gets only a moment's attention.
Also, there was a study about a year ago that concluded that some kinds of machine filtering have a lower false positive rate than humans.
-- Jamie
I get about 6k a day to my aging server, and the spam filtering cannot keep up. It's not an even trickle: some times of the day, it's several attempts per second. That's faster than the filtering software can handle.
Legitimate senders are getting "warning: could not send for past 4 hours" and then phoning me to ask if I've received their mail. The CPU and memory load spikes from time to time, and then it's not possible to login until it settles. (A known weak spot in Linux.) If I lower the resources allocated to mail processing, it cannot handle the incoming mail rate most of the time - it's just on the edge right now.
Now, I'm running Sendmail and Spamassassin on a Red Hat 9 box which is a 600MHz Celeron in a data centre. I'm sure all of you will laugh, and tell me to run better software on a newer OS and a better PC with a virtual machine. And how SpamFilter-of-the-day is much better than Spamassassin (it could hardly be worse).
But the fact is, updating all of those takes real time and expense. And when it's updated? It'll still be a significant load, and it'll still need to be maintained, and upgraded again at some point. I'm planning to order a server with 2GB of RAM because I think 1GB won't be enough to handle the memory load spikes for spam filtering.
Sorting all that will cost me real time and money. But at least I'll have reliable mail again and a server I can use for other things.
And that's nothing compared with the time I spend every fucking day skipping the spams in my inbox. About 5k per day are deleted by my filters. But that leaves 1k per day to skip manually - those where the risk of false positive is high enough that I need to check them. That's a pretty long and unpleasant inbox to face each day. Very unpleasant if I don't check it for a few days.
However, that has got to the point where _I'm_ accidentally deleting legitimate mails too. I have taken to whitelist-scoring my most important correspondents so that I don't accidently delete them among the spams. But that makes it even harder to respond to mail from people that have no prior connection with me, and people who haven't made it to the whitelist but should have.
Spam is definitely a financial problem for me. I estimate it costs me about $15000 USD/year in time spent deleting unwanted messages (about 40 minutes/day). Add on the hardware and software maintenance costs, and the annoyance, and the problems caused by deleting false positives.
For some definition of "full support". Let's see... I have Intel 945GM video in my laptop, and the latest Xorg drivers. (I have tried several experimental driver versions too).
Q: Does 3d acceleration work? A: Only on the laptop's internal screen. There is no 3d acceleration when using an external DVI screen - turning on acceleration causes video initialization to fail with an obscure should-never-happen error.
Q: Does dual display (laptop and external screen together) work? A: Only for some resolutions (i.e. not the actual size of the external TFT), and only with some driver versions and rather picky Xorg.conf settings (i.e. extremely flaky), and only with all hardware acceleration disabled and software cursor, and even when it works it's a most unpleasant and sometimes buggy hack (initializes whole video card twice; two driver instances both poking the same registers without synchronisation). The version I'm using at the moment (Ubuntu Edgy's "intel" driver) doesn't work with dual display, though it is supposed to. I'd use the other Intel driver ("i810") but it doesn't work on this laptop with a DVI external screen.
Q: Does simply displaying at a screen's native resolution work?! A: For modern TFT resolutions (mine are 1280x800 and 1680x1050): only with an experimental driver, or a BIOS hack (it temporarily modifies tables in the BIOS). Neither of these works with a laptop's internal and an external display simultaneously - you have to pick one when X starts, switching to the other requires X to be restarted, all applications closed etc.
Q: Does suspend work? A: No, after suspend-to-RAM, the screen is blank. (I've not tried suspend-to-disk because I use encrypted swap, and suspend-to-disk doesn't work with that.)
But to its credit, with the appropriate hacks and experimental drivers, displaying on one screen only with acceleration disabled and software cursor seems to be good... Not sure I'd call that "full support" though! I'm looking forward to it when it arrives. Intel have got the right idea, it's only implementation which is lacking now, and that appears to be outside Intel's control, being more of an under-resourced Xorg thing.
I was in a hotel in Shanghai last January, and although net access was good, I was never able to access the BBC site.
The thing is, we have a highly distributed database, the network is not constantly available, and updates (within constraints) may happen simultaneously at different places in the network and be reconciled later. And most of the nodes run on small embedded systems, while some nodes are large, centralised severs.
So many of the concepts you suggest are probably quite relevant theory, but we can't work using the common database implementations.
I'm not so concerned with how it's "mapped to physical tables". I regard that, since we're not using standard database engines, as storage and indexing optimisation. Important, but it's not really modelling.
I would like to read more about the things you mention, especially data modelling as a discipline, and the things you listed: "dynamic attributes", "snowflake schema" etc. Suggest any other good books?
Cheers,
-- Jamie
To be honest I'm working on a project where the data model is rather messy, and I designed it. But there is no _obvious_ right way to model it which fits with the ways we need to be able to access the data (there are real-time and distributed and offline aspects which mean we can't just use an RDB and standard normalisations).
Figuring out a good way forward would involve a couple of months break in visible features... and right now (as of a while), it's the features that are in constant demand. We're effectively writing a number of applications for different customers simultaneously, with most of the code in common, while we figure out what to do with insufficient resources to do the job 'properly'.
I suspect it's a common scenario.
And to be honest, we are servicing the business's short term needs quite well.
But I have to admit, the data model has plenty of room for improvement.
*rushes off to mock up a customer-specific feature...*
Will they start putting different content on each layer?
Extra scenes, only for licensed Blu-Ray users?
I can see this happening with data disks, to double the storage density if they're cheap enough to produce, and if they aren't superceded soon by another, higher-density format.
-- Jamie
On the contrary, modern CPUs have had RET-prediction for about 10 years now, and RET-prediction is almost always correct, so there's little or no pipeline penalty. It's better to have subroutines if they save cache space, than to flatten out code. That's why you see some routines being uninlined in the Linux kernel source, these days. There is still the overhead of setting up call arguments etc., so it's not always gainful. Register arguments help with that.
A practical difference between MIPS and ARM is that there are a number of MIPS clones which don't have to license anything from MIPS, provided they leave out the patented instructions. Whereas, cloning an ARM gets lots of hassle from ARM Ltd. Even just writing a software simulator for ARM is a problem(!) - this is why Qemu only emulates old ARMs.
So if someone's going to implement one of those instruction sets, it will tend to be MIPS if they're designing their own chip, and either of them if they're buying a pre-designed core.
No, often you can find an instruction to put after the branch. Sparc is the same. It's called the "branch delay slot", and it's a way of reducing the pipeline flush penalty.
Interesting, thanks!