Over the years I've owned modem/routers by Netgear, Linksys, Cisco, Swann, TP-Link, Thomson, Corega, Origo, D-Link, probably some others too. I've used them with different ISP in different countries, various physical locations, in the presence and absence of other local networks, with stock or custom firmware (DD-WRT, Tomato and OpenWRT) and I've come to one conclusion: they're all crap.
For the most part they work ok when new, but I have never had a router that stayed working for more than a couple of years. Usually the first symptom is wifi dropouts, then dropping ADSL, then router crashes. Eventually they just stop responding at all. It's not interference since getting a new router always fixes it all (for a bit anyway). I really don't understand how they would 'wear out' (being solid state); perhaps poor thermal design? I'm not a demanding user - they need to connect my ADSL to my ethernet and wifi quickly and reliably and provide NAT, DHCP, basic firewalling, operation as a DHCP relay or bridged access point if it's not a modem; I can happily do without media servers, port triggers, QoS, static routes, print serving etc - if it can't provide my most basic requirement of simply maintaining a connection, all that's just excess junk.
I've had the worst record with Netgear; their stuff just likes to die. I've not tried Apple, Buffalo or Draytek. Most of the open firmware is generally better than stock, but instability and risk of 'bricking' is stupidly high, and we wouldn't need them if stock firmware was decent in the first place.
It all seems a bit like my experience with Terry Pratchett books - I have a bad experience with one, so I look for suggestions, get recommended something, try that, and and am disappointed yet again. I'm certain there must be a market for wifi kit that isn't crap; it seems to be up for grabs.
There's a lot of conjecture and theorising in this thread so far. Not surprisingly some enterprising geeks have been busy testing SSDs to destruction, and they have some great stats. This thread with over 5000 posts has a ton of info about exactly what happens and some good hard numbers.
I'm not quite sure why, but in most of Europe the unit of choice is an inverse, typically litres per 100km, so lower figures are better; 54.5 miles per US gallon is 4.32 l/100km. A current, mainstream European large family estate car (station wagon, Peugeot 508 e-HDi / VW Passat Bluemotion) gets around 4.2 l/100km (56mpgUS). Peugeot also have a diesel-electric hybrid version that gets 3.2l/100km (61mpgUS). Curiously, Peugeot get some of their efficiency improvement via their automatic gearbox, which uses an automated manual system without a lossy torque converter; apparently it's crap though! VW's DSG gearbox is much better.
This on-amp argument is quite pathetic; European cars with way higher efficiency cope just fine on motorways with fewer lanes and higher speed limits (e.g. 83mph in France), i.e. where on-ramp aceleration is far more critical.
Ford's EcoBoost engine is pretty impressive. They have a 3-cyl, 1.0 litre version pushing out 177BHP at sane revs in the works. The V6 version is used in the F150 too.
This reminds me of a similar counter-argument for increasing car safety: if seatbelts were banned and cars were fitted with a large metal spike on the steering wheel, pointing towards the driver, you could be pretty sure that people would drive more carefully.
It could be that there is a correlation between poor riding practices and not wearing a helmet, given that not wearing a helmet is a poor practice!
They do give stats on helmet usage in non-fatal accidents: it's 13%. What you're saying about paying attention is true - accident prevention is by far the most effective measure - but the numbers make it clear that if you are going to have a crash, wearing a helmet improves your chances of surviving it by a factor of about 33, and you'd be hard pushed to find a factor that big by other means.
Where I usually ride I'd say helmet usage is in the high 90%s: It's very rare to see anyone without, and I don't know anyone that doesn't.
Cycling in London, I got used to using buses to my advantage - they're particularly useful as shields when pulling out onto roundabouts.
For a straightforward do-everything mail server (runs in an OpenBSD VM), take a look at mailserv. I'd really recommend weaning everyone off POP3 - it's just horrible. IMAP is great, and there's very little that doesn't support it now.
I've never found a contacts & calendaring solution with multi-user 2-way sync that really works - it always seems to run into trouble in one way or another. Any recommendations?
I've had real reliability problems with Google Apps, and their backup options are really quite bad: you can't seem to back up in a native format, only via conversions which are lossy, and as an admin, you can't back up user accounts - you need to log in to each one and back them up separately.
I did see this recently (asterisk/FreePBX running on a Raspberry Pi), it's got to be worth trying at the cost!
Helmets are certified to a minimum of 300g, and the reason for that figure is that it's been established (apparently) that it's roughly what you can take without sustaining significant injury. The helmet is thus a force measuring device of sorts - if it breaks it suggest that the impact exceeded 300g. If it breaks AND you did not sustain injury, it means that either the helmet was not up to standard (and the crash wasn't as bad as you thought), or that it provided sufficient critical protection while exceeding its failure threshold (i.e. it was a good helmet and saved your bacon). Random sample testing should reduce the incidence of the former (I don't know if that's common), and the latter is sustained by the fact that you're not dead - knowing precisely how much you're not dead by is not interesting and provides no rational basis for not wearing a helmet.
It's not that simple of course - some crashes will be unaccountably harsh (landing on a small stone on top of a hard surface wil send accelerations and force concentrations rocketing), others surprisingly mild.
I always wear one - I crashed into a pile of rocks at 40mph on the kamikaze downhill in Mammoth and had to dig 2" rock slivers out of my helmet (ok, and my arms, legs, neck and back too!) afterwards, but walked away mostly intact. A friend not wearing a helmet was killed with no external injuries from an 8mph fall onto flat pavement after being knocked off by a car.
It's not about total energy (which you can't do anything about), it's about how long you can draw out the impact time and thus reduce the acceleration you're subjected to: going from 2ms to 6ms is the difference between life and death - helmets absolutely ARE crumple zones; disintegration is a failure mode when they have nothing left to give.
Of course you can't do that, but you could do reasonable tests on a helmet of the same type to show that it might require a certain amount of acceleration to break, and it would show whether exceeding its failure threshold is still likely to bring an impact to within a survivable range. All recent helmet certifications have a minimum failure limit of 300g, so it should take more than that to actually break, but since an unprotected impact is likely to exceed 1000g it's not unreasonable to say that a broken helmet could well have saved you, perhaps by lowering the impact to 400g - that's not a good number, but it's quite definitely less bad. Anything that increases impact time (and thus reduces acceleration) is a good thing.
Right, because pedestrians ordinarily share their designated space with fast-moving, 2-tonne chunks of metal with limited maneuverability and inattentive operators. It's not your speed that matters. A friend (not wearing a helmet) was killed while doing ~8mph (up a hill) when he was knocked off his bike by a car that turned across him. No broken bones, no blood, just a fatal haematoma that would probably have been prevented by a helmet (or by not getting hit).
Nonsense. Helmets reduce acceleration on impact, and that applies to intra-cranial effects too. Without a helmet, a 12mph fall from 2m is likely to result in a head acceleration of over 1000g in the 2ms or so it takes to stop; helmets are designed to keep that below 300g (often less) by extending the acceleration time to 6-8ms.
In London, the average speed of road traffic is 13mph, dropping to 7mph at peak time. Bicycles are usually the fastest thing on the roads. In big cities, cars are the problem, not the solution.
You've missed one of TrueCrypt's major features: plausible deniability. You can provide keys that unlock encrypted data on your HD. It may not reveal all the data, but it's not possible for your accuser to prove that it's not.
MySQLs handlersocket (included since 5.5) does NoSQL-style read and write operations bypassing the SQL engine. While it has some limitations, it will do >200,000 queries/sec on a low-spec server and there are benchmarks of it doing >750,000 on a 8-core Nehalem (faster than Memcached!), and it's not restricted to in-memory operations. The nice thing is that you can use that for the simpler parts of your app, then use transactional SQL on the same database for more complex operations.
Another one to look at is TokuTek's TokuDB, another InnoDB drop-in replacement, which is particularly good for inserts, low disk use and low-latency replication. They ran a demo doing 1 billion indexed inserts in 7 hours when InnoDB took a week.
For distributed 'cloudy' apps, one of the better choices is Drizzle, which retains the nice bits of MySQL (and MySQL client compatibility) and rewrites all the rest.
I don't think I'll believe MemSQL until Percona have benchmarked it...
There are plenty of earth models for climatic and other purposes. It's clearly not practical to make physical models, so we have to make do with software ones which don't have such practical constraints. Their accuracy can be tested by seeing if older data can be used to predict more recent data (hindcasting), for example can data gathered from 1900 to 1960 in a given model be used to predict what the conditions were like in the 1960s? If they do, then you might consider some of that model's future predictions trustworthy too. This technique is used to test models of individual parts of an overall climate model, such as temperature changes, cloud actions, El Niño events, gas mixtures etc. Generally these models will only ever get better as research improves and computing power increases. Still, they are an approximation (as all models necessarily are), but as the IPCC said: "Despite such uncertainties, however, models are unanimous in their prediction of substantial climate warming under greenhouse gas increases". More info.
I was just reading about that today. Since issuing loans with fractional reserve is more or less a zero-cost for the bank (apart from a minor admin overhead unrelated to the size of the loan), interest isn't really deserved. So what would happen if interest wasn't payable on loans with no reserve behind them, or perhaps fractionally according to the fraction of reserve? Obviously it would make the loans far less attractive to the banks, but would it also undermine the markets built on the virtual money printing operation they represent (which isn't much to do with the interest)?
Slashdot Mirror
There are a variety of hypothesized ways to generate entirely new species, but they have not been observed directly.
Not true.
CREATE TABLE `people` ( `name` VARCHAR(255), `religion` ENUM('Christian', 'Muslim', 'Hindu', 'Other', '') DEFAULT '' );
Over the years I've owned modem/routers by Netgear, Linksys, Cisco, Swann, TP-Link, Thomson, Corega, Origo, D-Link, probably some others too. I've used them with different ISP in different countries, various physical locations, in the presence and absence of other local networks, with stock or custom firmware (DD-WRT, Tomato and OpenWRT) and I've come to one conclusion: they're all crap.
For the most part they work ok when new, but I have never had a router that stayed working for more than a couple of years. Usually the first symptom is wifi dropouts, then dropping ADSL, then router crashes. Eventually they just stop responding at all. It's not interference since getting a new router always fixes it all (for a bit anyway). I really don't understand how they would 'wear out' (being solid state); perhaps poor thermal design? I'm not a demanding user - they need to connect my ADSL to my ethernet and wifi quickly and reliably and provide NAT, DHCP, basic firewalling, operation as a DHCP relay or bridged access point if it's not a modem; I can happily do without media servers, port triggers, QoS, static routes, print serving etc - if it can't provide my most basic requirement of simply maintaining a connection, all that's just excess junk.
I've had the worst record with Netgear; their stuff just likes to die. I've not tried Apple, Buffalo or Draytek. Most of the open firmware is generally better than stock, but instability and risk of 'bricking' is stupidly high, and we wouldn't need them if stock firmware was decent in the first place.
It all seems a bit like my experience with Terry Pratchett books - I have a bad experience with one, so I look for suggestions, get recommended something, try that, and and am disappointed yet again. I'm certain there must be a market for wifi kit that isn't crap; it seems to be up for grabs.
There's a lot of conjecture and theorising in this thread so far. Not surprisingly some enterprising geeks have been busy testing SSDs to destruction, and they have some great stats. This thread with over 5000 posts has a ton of info about exactly what happens and some good hard numbers.
I'm not quite sure why, but in most of Europe the unit of choice is an inverse, typically litres per 100km, so lower figures are better; 54.5 miles per US gallon is 4.32 l/100km. A current, mainstream European large family estate car (station wagon, Peugeot 508 e-HDi / VW Passat Bluemotion) gets around 4.2 l/100km (56mpgUS). Peugeot also have a diesel-electric hybrid version that gets 3.2l/100km (61mpgUS). Curiously, Peugeot get some of their efficiency improvement via their automatic gearbox, which uses an automated manual system without a lossy torque converter; apparently it's crap though! VW's DSG gearbox is much better.
This on-amp argument is quite pathetic; European cars with way higher efficiency cope just fine on motorways with fewer lanes and higher speed limits (e.g. 83mph in France), i.e. where on-ramp aceleration is far more critical.
Ford's EcoBoost engine is pretty impressive. They have a 3-cyl, 1.0 litre version pushing out 177BHP at sane revs in the works. The V6 version is used in the F150 too.
This reminds me of a similar counter-argument for increasing car safety: if seatbelts were banned and cars were fitted with a large metal spike on the steering wheel, pointing towards the driver, you could be pretty sure that people would drive more carefully.
It could be that there is a correlation between poor riding practices and not wearing a helmet, given that not wearing a helmet is a poor practice!
They do give stats on helmet usage in non-fatal accidents: it's 13%. What you're saying about paying attention is true - accident prevention is by far the most effective measure - but the numbers make it clear that if you are going to have a crash, wearing a helmet improves your chances of surviving it by a factor of about 33, and you'd be hard pushed to find a factor that big by other means.
Where I usually ride I'd say helmet usage is in the high 90%s: It's very rare to see anyone without, and I don't know anyone that doesn't.
Cycling in London, I got used to using buses to my advantage - they're particularly useful as shields when pulling out onto roundabouts.
For a straightforward do-everything mail server (runs in an OpenBSD VM), take a look at mailserv. I'd really recommend weaning everyone off POP3 - it's just horrible. IMAP is great, and there's very little that doesn't support it now.
I've never found a contacts & calendaring solution with multi-user 2-way sync that really works - it always seems to run into trouble in one way or another. Any recommendations?
I've had real reliability problems with Google Apps, and their backup options are really quite bad: you can't seem to back up in a native format, only via conversions which are lossy, and as an admin, you can't back up user accounts - you need to log in to each one and back them up separately.
I did see this recently (asterisk/FreePBX running on a Raspberry Pi), it's got to be worth trying at the cost!
There's also Percona server, which is very well supported.
Helmets are certified to a minimum of 300g, and the reason for that figure is that it's been established (apparently) that it's roughly what you can take without sustaining significant injury. The helmet is thus a force measuring device of sorts - if it breaks it suggest that the impact exceeded 300g. If it breaks AND you did not sustain injury, it means that either the helmet was not up to standard (and the crash wasn't as bad as you thought), or that it provided sufficient critical protection while exceeding its failure threshold (i.e. it was a good helmet and saved your bacon). Random sample testing should reduce the incidence of the former (I don't know if that's common), and the latter is sustained by the fact that you're not dead - knowing precisely how much you're not dead by is not interesting and provides no rational basis for not wearing a helmet.
It's not that simple of course - some crashes will be unaccountably harsh (landing on a small stone on top of a hard surface wil send accelerations and force concentrations rocketing), others surprisingly mild.
I always wear one - I crashed into a pile of rocks at 40mph on the kamikaze downhill in Mammoth and had to dig 2" rock slivers out of my helmet (ok, and my arms, legs, neck and back too!) afterwards, but walked away mostly intact. A friend not wearing a helmet was killed with no external injuries from an 8mph fall onto flat pavement after being knocked off by a car.
It's not about total energy (which you can't do anything about), it's about how long you can draw out the impact time and thus reduce the acceleration you're subjected to: going from 2ms to 6ms is the difference between life and death - helmets absolutely ARE crumple zones; disintegration is a failure mode when they have nothing left to give.
Of course you can't do that, but you could do reasonable tests on a helmet of the same type to show that it might require a certain amount of acceleration to break, and it would show whether exceeding its failure threshold is still likely to bring an impact to within a survivable range. All recent helmet certifications have a minimum failure limit of 300g, so it should take more than that to actually break, but since an unprotected impact is likely to exceed 1000g it's not unreasonable to say that a broken helmet could well have saved you, perhaps by lowering the impact to 400g - that's not a good number, but it's quite definitely less bad. Anything that increases impact time (and thus reduces acceleration) is a good thing.
Right, because pedestrians ordinarily share their designated space with fast-moving, 2-tonne chunks of metal with limited maneuverability and inattentive operators. It's not your speed that matters. A friend (not wearing a helmet) was killed while doing ~8mph (up a hill) when he was knocked off his bike by a car that turned across him. No broken bones, no blood, just a fatal haematoma that would probably have been prevented by a helmet (or by not getting hit).
97% of cyclists killed in New York between 1996 and 2005 were not wearing helmets.
A friend of mine was stopped and cautioned by the police for cycling at 40mph (downhill) in a 30 limit.
Nonsense. Helmets reduce acceleration on impact, and that applies to intra-cranial effects too. Without a helmet, a 12mph fall from 2m is likely to result in a head acceleration of over 1000g in the 2ms or so it takes to stop; helmets are designed to keep that below 300g (often less) by extending the acceleration time to 6-8ms.
In London, the average speed of road traffic is 13mph, dropping to 7mph at peak time. Bicycles are usually the fastest thing on the roads. In big cities, cars are the problem, not the solution.
You've missed one of TrueCrypt's major features: plausible deniability. You can provide keys that unlock encrypted data on your HD. It may not reveal all the data, but it's not possible for your accuser to prove that it's not.
MySQLs handlersocket (included since 5.5) does NoSQL-style read and write operations bypassing the SQL engine. While it has some limitations, it will do >200,000 queries/sec on a low-spec server and there are benchmarks of it doing >750,000 on a 8-core Nehalem (faster than Memcached!), and it's not restricted to in-memory operations. The nice thing is that you can use that for the simpler parts of your app, then use transactional SQL on the same database for more complex operations.
Another one to look at is TokuTek's TokuDB, another InnoDB drop-in replacement, which is particularly good for inserts, low disk use and low-latency replication. They ran a demo doing 1 billion indexed inserts in 7 hours when InnoDB took a week.
For distributed 'cloudy' apps, one of the better choices is Drizzle, which retains the nice bits of MySQL (and MySQL client compatibility) and rewrites all the rest.
I don't think I'll believe MemSQL until Percona have benchmarked it...
> As far as I know no one has created a model of the earth to test global warming or bred a large number of animals to create a new species.
No-one that is, apart from those that have. There have been a fairly large number of the latter, both observing and inducing speciation in plants and animals.
There are plenty of earth models for climatic and other purposes. It's clearly not practical to make physical models, so we have to make do with software ones which don't have such practical constraints. Their accuracy can be tested by seeing if older data can be used to predict more recent data (hindcasting), for example can data gathered from 1900 to 1960 in a given model be used to predict what the conditions were like in the 1960s? If they do, then you might consider some of that model's future predictions trustworthy too. This technique is used to test models of individual parts of an overall climate model, such as temperature changes, cloud actions, El Niño events, gas mixtures etc. Generally these models will only ever get better as research improves and computing power increases. Still, they are an approximation (as all models necessarily are), but as the IPCC said: "Despite such uncertainties, however, models are unanimous in their prediction of substantial climate warming under greenhouse gas increases". More info.
The 'Apple' key's official name is 'command', so yes.
Not encountered ctrl-cmd-D?
OS X has BSD roots and is an officially certified UNIX. Linux is not.
I was just reading about that today. Since issuing loans with fractional reserve is more or less a zero-cost for the bank (apart from a minor admin overhead unrelated to the size of the loan), interest isn't really deserved. So what would happen if interest wasn't payable on loans with no reserve behind them, or perhaps fractionally according to the fraction of reserve? Obviously it would make the loans far less attractive to the banks, but would it also undermine the markets built on the virtual money printing operation they represent (which isn't much to do with the interest)?