Well, you could put your order in any time before deadline, but orders are only executed at the instant of the deadline. So, you get all the "thinking" time you need, and once the deadline is reached, orders are processed simultaneously (or in random order). This is similar to how concert tickets (or, for that matter, new issues of stock) are sold: you have a week to post in your cheque, and then all envelopes are opened at the same time - in the interests of fairness, the should be shuffled first if there is a risk that not all orders can be fulfilled.
Actually, I think we can draw a line. It takes about 200ms for an electrical pulse to travel round the world (speed of light in glass is lower than c), and we have a bit of switching delay. So this should imply the minimum timing limit. Anyway, fortunately the exchange can set the rules here, if it wants to.
Honestly, this is really a bad idea for overall market stability. What we really need is a much slower, yet fairer system.
What I'd suggest is something similar to synchronous clocking:
Every second, on the second, prices are published.
500 ms later, orders are placed and fulfilled.
500 ms later, the updated prices are published.
Benefit #1: fairness - those who are closest to the exchange or have stupidly fast hardware can't get in front of the rest. Benefit #2: slower responses. If the clock can only "tick" 60 times a minute, there is a chance for human intervention before disasters happen. Benefit #3: markets are more able to serve the rest of society, rather than being used purely for "gambling". imho, the existence of "high frequency trading" is a kind of tragedy of the commons: nobody really "wins", but if everyone else does it, and you don't, you lose.
Actually, it wasn't open-source enough. Many of the nasty bugs in the platform related to closed-source components. For example, I wanted to be able to simultaneously run a VoIP call and use the camera. But the camera "helpfully" wouldn't run during a call, because it needed the sound card to make the "click" noise. I tried to fix it, but was told that the camera app was closed source. There was an open-source camera library component...but only if I didn't want auto-focus! Another N900 killer was video-calling. Almost there, but pulse-audio was hogging 30% CPU doing (un-needed) sample-rate conversion - and we couldn't remove it.
Worth pointing out that, if you do 20 hypothesis tests in a study, all at the 5% level, your expectation should be that approximately 1 of your conclusions is false. Also, between subjective errors, regression to the mean, and publication bias, it's not surprising that at least some of these major results turn out to be wrong....
Why shouldn't they release a Linux client with DRM? They surely aren't obliged to only support unbroken DRM. Ideally they'd release an open source client (with DRM) that someone else would then compile out - but even a binary blob that runs under Wine would suffice. As always, the Free Software community is likely to be enthusiastic to do the work, if only it is provided with some minimal assistance (such as a spec and documentation).
- Cover the wall in network and power-points - you might well make use of 12 ethernet ports and 50 mains sockets. - Have deep, load-bearing shelves above the workbench, (again, with power), and under-shelf lighting. - Consider the ergonomics: workbench height for standing (and some tall stools), a/c, bright light, silent computers. LCD monitor on a swing-arm? - If money is limited, you're probably better off with a larger variety of stock and tools than with fewer expensive ones. - Do you need the ability to make it dark? Plumbing? Dust/Fume extraction? - Ensure the floor is easy to clean, not static-prone, and easy to see where you dropped things.
Also, if Nokia really wants to succeed, it has to make the *entire* platform open, rather than 3/4 open with all the interesting bits closed. I bought an N900, and it was a delight to develop for, particularly because it has the good sense to run X-windows, so standard Linux apps can easily be ported (even X-forwarding over SSH worked!). BUT there was a killer bug(*) with the camera, which I couldn't fix, since the autofocus algorithm was only available to the binary-blobs. As a result, we didn't buy 400 of them.
(*)Specifically, you can't use the camera while in the middle of a phone-call, because the camera wants to make a shutter-release noise, but the sound interface is locked because the phone is using it. The simple workaround of taking the photo without a shutter-release click couldn't be done because Nokia wouldn't release the source. I think they fixed it now, but months too late for me. Even more ridiculous was the fact that they did support using the camera in your own software...but only if you didn't want it to have autofocus!
If you can justify it, go for mini-itx boards with SSDs. You'll pay about $300-400 per machine.
Advantages:
- Very much reduced power consumption. Electricity isn't cheap anymore - consider $1/week for a desktop.
- Silent. (You really appreciate this when you get used to it). No moving parts => less to break.
- SSDs are really really fast. Even a value 64GB SSD, coupled with 4 GB RAM will make the machine fly, and
the D525 CPU is not at all bad (it's effectively a quad-core)
- Very easy to build
- Mount the PC on the back of the monitor for compactness. A VGA mount should cost only $10 (I had some made out
of aluminium angle for $5)
- Use the saving to buy 24" LCDs for everyone - this is a major productivity win.
If you really can't live with an Atom CPU, (though the D525 is a 64-bit quad-core!), you can get mini-itx boxes that take Intel Core2 (or similar) CPUs, though these do have fans.
Regarding software, turn off all the flashy desktop animations and the machine will appear much snappier. Also, consider whether Etherboot/Linux has anything to offer you - even if you just use a customised Ubuntu install usb-key in order to get the machine automatically set-up with your Windows environment. (Or perhaps every night, send a wake-on-lan packet, etherboot, and rsync the image across to update).
Lastly, go for one of the Mini-itx boards with its own internal 12V power-supply on the motherboard (uses a laptop-style brick) - this eliminates another step from assembly. You should be able to get the Mitx systems fully assembled in half-an-hour each, once you've built the first 10 (though you can get suppliers who will build them for you).
Make sure the employees like them: buy good keyboards and mice, + the gel wrist-rests. (Personally, I really like IBM ultranav keyboards)
One way to do this would be for the android market to list (or provide a filter for) how the app developer makes his money. Of course I prefer GPL apps, but I'm willing to install free (beer) apps, trial versions (if clearly indicated), or paid apps. However, I don't want to have to install an app just to discover that it's adware.
I wonder whether the synchronous-counter approach would help reduce glitches here. In other words:
- at HH:MM:00 update the prices (and allow the change to propagate; everyone can put in their next trade order)
- at HH:MM:30 execute the trades (and then there are 30 seconds to decide on the new price before a value is propagated)
This is the way that synchronous logic works. The current model is more like hundreds of ripple-counters.
(It would also ensure slightly greater fairness by not giving an advantage to the person with the absolutely fastest network connection, and would slow down the cycles so that a market collapse took many minutes.)
Why not let people choose whether to be protected *with* patents, or to be protected *from* patents. So big companies can file patents and sue each other; small companies who choose not to file can neither sue, nor be sued.
Did you know that, in recent Firefoxes, Ctrl-+ will zoom the entire page, not just magnify the text? This will fix most of your problems. Sadly, flash is the one dishonourable exception, though you could consider a screen magnifier utility if it gets really bad.
So...when is google going to implement the iPhone API on Android? If the Wine project can make MS apps run on Linux, then why not have an "iWine" project to make iPhone apps run elsewhere?
If you exclude the TxD and RxD (and Ground) lines, there are still 6 others, which are trivial to interface to. YOu get 4 inputs and 2 outputs, to which you can connect switches and LEDs directly. Then use setserial/statserial to control/monitor the logic levels. This is actually quite useful sometimes.
We're also very happy with it. I have a RAIS array of two servers each with RAID1. So our Postgres database is configured on a quadruple-backup setup thus:
Postgres/var/lib/postgresql/dev/drbd1 primary server --- secondary server raid 1 raid 1 2 x X25-E SSD 2 x X25-E SSD
The servers are connected back to back by a direct gigabit ethernet link, and we use DRBD in protocol B (memory synchronous). Thus all transactions are guaranteed to hit the disk, we get fast performance, and excellent reliability. Cutover from one machine to the other is really easy, and takes 2 minutes. The docs for DRBD are also very good.
There are 2 things to note:
* Write bandwidth this way is about half what it could be with a single server.
* Avoid the risk of "split-brain" at all costs. For us, this means a manual failover process: dataloss is very bad, but a few minutes of downtime is acceptable.
Well, I may not be able to control my identity, but I can repudiate my biometrics. The idea would be (if enough people participated) that a given fingerprint or DNA sample at a crime scene etc would cease to identify any one person in any reliable way.
How about a public (anonymised) repository of fingerprints. The idea is this: I can't change my prints, nor can I get back control once the government has taken them. But I could publish them to the world. That makes the print very easy for anyone else to fake. In other words, plausible deniability.
If you compare the best-of-breed tungsten with the best-of-breed CFL, the factor is about 3.3. That factor of 5 (I remember when it used to be 7) is derived from a comparison with a rather poor tungsten bulb (eg 40W, which are far less efficient than 100W).
I looked at what's available now in the shops:
Best Tungsten: 105 W in, 2100 lumen out.(*)
Best CFL: 18W in, 1200 lumen out.
Efficiency difference: factor of 3.3
(*) This is the Philips "Halogen Energy Saver -30%", which also lasts 2000 hours. I'm writing this in the UK; if it were the USA, the lower mains voltage acts even more in Tungsten's favour.
In addition, Tungsten bulbs:
* Give a much better light: full spectrum, better colours, warmer, more flattering to females, and not cold/depressing in wintertime.
* Run at (virtually) full efficiency from start to end; whereas CFLs need to be replaced after 6 months as they age.
* Give out waste heat, but this is still offset against the heat you'd otherwise burn natural gas to get; electricity is potentially green (Nuclear), whereas gas never is; also the waste heat is usually in the same room that you currently occupy, rather than around the entire house.
* Run at a power factor of 1.0, whereas CFLs are inductive, wasting a lot of energy in transmission and generation. (the consumer isn't directly metered for this, but it pushes up the price of electricty in general).
* Could be improved further if we combine the halogen technology (in the Philips bulbs) with the Dichroic technology (used by Osram) to reflect IR back to the filament.
So, yes, there is a slight figure of merit for fluorescent bulbs, but it's pretty small.
Well, data = information = entropy. What's the minimum amount of energy required (in theory) to store a single bit of data? It's probably not zero. Therefore data does actually have energy, i.e. mass.
As a UK citizen, who used to come to the US on holiday twice a year (Florida in the summer; Colorado for the skiing), we've transferred our vacations to other places (usually Italy in summer; Canada for skiing). The main reasons are Guantanamo, and the fingerprinting process. In the last 8 years, our family has cost the US economy $100k+ as a result. When will it stop?
Slashdot Mirror
Well, you could put your order in any time before deadline, but orders are only executed at the instant of the deadline. So, you get all the "thinking" time you need, and once the deadline is reached, orders are processed simultaneously (or in random order). This is similar to how concert tickets (or, for that matter, new issues of stock) are sold: you have a week to post in your cheque, and then all envelopes are opened at the same time - in the interests of fairness, the should be shuffled first if there is a risk that not all orders can be fulfilled.
Actually, I think we can draw a line. It takes about 200ms for an electrical pulse to travel round the world (speed of light in glass is lower than c), and we have a bit of switching delay. So this should imply the minimum timing limit. Anyway, fortunately the exchange can set the rules here, if it wants to.
Honestly, this is really a bad idea for overall market stability. What we really need is a much slower, yet fairer system.
What I'd suggest is something similar to synchronous clocking:
Every second, on the second, prices are published.
500 ms later, orders are placed and fulfilled.
500 ms later, the updated prices are published.
Benefit #1: fairness - those who are closest to the exchange or have stupidly fast hardware can't get in front of the rest.
Benefit #2: slower responses. If the clock can only "tick" 60 times a minute, there is a chance for human intervention before disasters happen.
Benefit #3: markets are more able to serve the rest of society, rather than being used purely for "gambling". imho, the existence of "high frequency trading" is a kind of tragedy of the commons: nobody really "wins", but if everyone else does it, and you don't, you lose.
It may be a stupid idea, but presumably the patient gave his informed consent. Much cosmetic surgery isn't medically necessary either.
Actually, it wasn't open-source enough. Many of the nasty bugs in the platform related to closed-source components. For example, I wanted to be able to simultaneously run a VoIP call and use the camera. But the camera "helpfully" wouldn't run during a call, because it needed the sound card to make the "click" noise. I tried to fix it, but was told that the camera app was closed source. There was an open-source camera library component...but only if I didn't want auto-focus! Another N900 killer was video-calling. Almost there, but pulse-audio was hogging 30% CPU doing (un-needed) sample-rate conversion - and we couldn't remove it.
...one could patent a step necessary in obtaining patents, and then refuse to license it to anybody.
Worth pointing out that, if you do 20 hypothesis tests in a study, all at the 5% level, your expectation should be that approximately 1 of your conclusions is false.
Also, between subjective errors, regression to the mean, and publication bias, it's not surprising that at least some of these major results turn out to be wrong....
Why shouldn't they release a Linux client with DRM? They surely aren't obliged to only support unbroken DRM. Ideally they'd release an open source client (with DRM) that someone else would then compile out - but even a binary blob that runs under Wine would suffice. As always, the Free Software community is likely to be enthusiastic to do the work, if only it is provided with some minimal assistance (such as a spec and documentation).
- Cover the wall in network and power-points - you might well make use of 12 ethernet ports and 50 mains sockets.
- Have deep, load-bearing shelves above the workbench, (again, with power), and under-shelf lighting.
- Consider the ergonomics: workbench height for standing (and some tall stools), a/c, bright light, silent computers. LCD monitor on a swing-arm?
- If money is limited, you're probably better off with a larger variety of stock and tools than with fewer expensive ones.
- Do you need the ability to make it dark? Plumbing? Dust/Fume extraction?
- Ensure the floor is easy to clean, not static-prone, and easy to see where you dropped things.
Also, if Nokia really wants to succeed, it has to make the *entire* platform open, rather than 3/4 open with all the interesting bits closed. I bought an N900, and it was a delight to develop for, particularly because it has the good sense to run X-windows, so standard Linux apps can easily be ported (even X-forwarding over SSH worked!). BUT there was a killer bug(*) with the camera, which I couldn't fix, since the autofocus algorithm was only available to the binary-blobs. As a result, we didn't buy 400 of them.
(*)Specifically, you can't use the camera while in the middle of a phone-call, because the camera wants to make a shutter-release noise, but the sound interface is locked because the phone is using it. The simple workaround of taking the photo without a shutter-release click couldn't be done because Nokia wouldn't release the source. I think they fixed it now, but months too late for me. Even more ridiculous was the fact that they did support using the camera in your own software...but only if you didn't want it to have autofocus!
If you can justify it, go for mini-itx boards with SSDs. You'll pay about $300-400 per machine.
Advantages:
- Very much reduced power consumption. Electricity isn't cheap anymore - consider $1/week for a desktop.
- Silent. (You really appreciate this when you get used to it). No moving parts => less to break.
- SSDs are really really fast. Even a value 64GB SSD, coupled with 4 GB RAM will make the machine fly, and
the D525 CPU is not at all bad (it's effectively a quad-core)
- Very easy to build
- Mount the PC on the back of the monitor for compactness. A VGA mount should cost only $10 (I had some made out
of aluminium angle for $5)
- Use the saving to buy 24" LCDs for everyone - this is a major productivity win.
If you really can't live with an Atom CPU, (though the D525 is a 64-bit quad-core!), you can get mini-itx boxes that take Intel Core2 (or similar) CPUs, though these do have fans.
Regarding software, turn off all the flashy desktop animations and the machine will appear much snappier. Also, consider whether Etherboot/Linux has anything to offer you - even if you just use a customised Ubuntu install usb-key in order to get the machine automatically set-up with your Windows environment. (Or perhaps every night, send a wake-on-lan packet, etherboot, and rsync the image across to update).
Lastly, go for one of the Mini-itx boards with its own internal 12V power-supply on the motherboard (uses a laptop-style brick) - this eliminates another step from assembly. You should be able to get the Mitx systems fully assembled in half-an-hour each, once you've built the first 10 (though you can get suppliers who will build them for you).
Make sure the employees like them: buy good keyboards and mice, + the gel wrist-rests. (Personally, I really like IBM ultranav keyboards)
One way to do this would be for the android market to list (or provide a filter for) how the app developer makes his money.
Of course I prefer GPL apps, but I'm willing to install free (beer) apps, trial versions (if clearly indicated), or paid apps.
However, I don't want to have to install an app just to discover that it's adware.
I only ever use the 5.25" bays, and find the 3.5" ones useless.
For example:
http://www.quietpc.com/gb-en-gbp/products/harddrivesolutions/smartdrive-neo
takes a conventional 3.5" disk, and both cools and quietens it.
The effect is quite dramatic: I cannot hear my PC at all.
(I already put a silent CPU heatsink in, and a very-quiet PSU).
I wonder whether the synchronous-counter approach would help reduce glitches here. In other words:
- at HH:MM:00 update the prices (and allow the change to propagate; everyone can put in their next trade order)
- at HH:MM:30 execute the trades (and then there are 30 seconds to decide on the new price before a value is propagated)
This is the way that synchronous logic works. The current model is more like hundreds of ripple-counters.
(It would also ensure slightly greater fairness by not giving an advantage to the person with the absolutely fastest network connection, and would slow down the cycles so that a market collapse took many minutes.)
Why not let people choose whether to be protected *with* patents, or to be protected *from* patents.
So big companies can file patents and sue each other; small companies who choose not to file can neither sue, nor be sued.
Did you know that, in recent Firefoxes, Ctrl-+ will zoom the entire page, not just magnify the text?
This will fix most of your problems. Sadly, flash is the one dishonourable exception, though you could
consider a screen magnifier utility if it gets really bad.
So...when is google going to implement the iPhone API on Android? If the Wine project can make MS apps run on Linux, then why not have an "iWine" project to make iPhone apps run elsewhere?
If you exclude the TxD and RxD (and Ground) lines, there are still 6 others, which are trivial to interface to. YOu get 4 inputs and 2 outputs, to which you can connect switches and LEDs directly. Then use setserial/statserial to control/monitor the logic levels. This is actually quite useful sometimes.
We're also very happy with it. I have a RAIS array of two servers each with RAID1. So our Postgres database is configured on a quadruple-backup setup thus:
Postgres /var/lib/postgresql /dev/drbd1
primary server --- secondary server
raid 1 raid 1
2 x X25-E SSD 2 x X25-E SSD
The servers are connected back to back by a direct gigabit ethernet link, and we use DRBD in protocol B (memory synchronous).
Thus all transactions are guaranteed to hit the disk, we get fast performance, and excellent reliability.
Cutover from one machine to the other is really easy, and takes 2 minutes. The docs for DRBD are also very good.
There are 2 things to note:
* Write bandwidth this way is about half what it could be with a single server.
* Avoid the risk of "split-brain" at all costs. For us, this means a manual failover process: dataloss is very bad, but a few minutes of downtime is acceptable.
Well, I may not be able to control my identity, but I can repudiate my biometrics. The idea would be (if enough people participated) that a given fingerprint or DNA sample at a crime scene etc would cease to identify any one person in any reliable way.
How about a public (anonymised) repository of fingerprints. The idea is this: I can't change my prints, nor can I get back control once the government has taken them. But I could publish them to the world. That makes the print very easy for anyone else to fake. In other words, plausible deniability.
If you compare the best-of-breed tungsten with the best-of-breed CFL, the factor is about 3.3. That factor of 5 (I remember when it used to be 7) is derived from a comparison with a rather poor tungsten bulb (eg 40W, which are far less efficient than 100W).
I looked at what's available now in the shops:
Best Tungsten: 105 W in, 2100 lumen out.(*)
Best CFL: 18W in, 1200 lumen out.
Efficiency difference: factor of 3.3
(*) This is the Philips "Halogen Energy Saver -30%", which also lasts 2000 hours. I'm writing this in the UK; if it were the USA, the lower mains voltage acts even more in Tungsten's favour.
In addition, Tungsten bulbs:
* Give a much better light: full spectrum, better colours, warmer, more flattering to females, and not cold/depressing in wintertime.
* Run at (virtually) full efficiency from start to end; whereas CFLs need to be replaced after 6 months as they age.
* Give out waste heat, but this is still offset against the heat you'd otherwise burn natural gas to get; electricity is potentially
green (Nuclear), whereas gas never is; also the waste heat is usually in the same room that you currently occupy, rather than
around the entire house.
* Run at a power factor of 1.0, whereas CFLs are inductive, wasting a lot of energy in transmission and generation. (the consumer isn't directly metered for this, but it pushes up the price of electricty in general).
* Could be improved further if we combine the halogen technology (in the Philips bulbs) with the Dichroic technology (used by Osram) to reflect IR back to the filament.
So, yes, there is a slight figure of merit for fluorescent bulbs, but it's pretty small.
Well, that's the question. Is it really so? (This is an interesting quantum-mechanical conundrum, to which I don't know the answer).
Well, data = information = entropy. What's the minimum amount of energy required (in theory) to store a single bit of data?
It's probably not zero. Therefore data does actually have energy, i.e. mass.
As a UK citizen, who used to come to the US on holiday twice a year (Florida in the summer; Colorado for the skiing), we've transferred our vacations to other places (usually Italy in summer; Canada for skiing). The main reasons are Guantanamo, and the fingerprinting process. In the last 8 years, our family has cost the US economy $100k+ as a result. When will it stop?