Other projects, as requested: Bifferboard [bifferos.co.uk] and Beagleboard [beagleboard.org]. Both longer-established and actually in production. Bifferboard costs the same as Raspberry Pi Model B.
Essentially with the Pi you are getting a device with specs comparable to the beableboard but a price comparable to (probablly lower than but we won't know for sure until we see final shipped prices from the Pi guys) the bifferboard.
According to the actual proposal they plan to support them for 54 weeks with an overlap of 12 weeks.
That is very little overlap between releases in which to plan your upgrade policy and since the changes won't come at any particular time of year it will be difficult to tie it in with other upgrades.
The problem comes when the import duty for the constituent parts/materials works out more expensive than the import duty for what is being manufactured. Then you have a disincentive to manufacture locally.
IMO a sane import duty system would have relatively high import duties for finished goods and lower import duties the further back along the production chain you moved so the incentive was to do as much of the manufacturing as possible in-country but afaict duties are set to pander to specific interests so they often don't end up sane.
Another problem is that many countries have exemptions for small imports. Even if they don't it's much easier to get away with dodgy declarations on small imports. This make sense from the point of view of saving paperwork (often the charge for collecting import duty/VAT is more than the duty/VAT itself) but it also creates an incentive for customers to buy directly from abroad.
If you don't know what your computer is doing and why it would be writing to a removable device then that's your own failing.
I'm not the original poster and i've never had issues with removable media that I couldn't find an application to blame for (some apps seem to keep files open even after the user has "closed" the visible representation of the file in the program). I'd bet very few users really know what is running in the background on their computer and what it is doing. Windows is worse because apps tend to be closed source and not filtered by a distro but even on linux do you really know the exact details of how every program you have installed behaves?
And afaict a file open for read is more than enough to prevent unmouning a partition on both windows and linux. It doesn't need to be writing it.
Linux does that just the same too - ever seen 'Device or resource busy'? No help there. Allowing you to force unmount is not a safe thing to do for casual users - you need to know what you're doing and be prepared to accept the consequences of a mistake.
If the OS lets the user force unmount then at least the FS gets the opertunity to shut down cleanly even if the app doesn't. If the OS doesn't let the user force unmount and the user pulls the plug anyway then neither the FS or the app gets to shut down cleanly.
Though as you say linux isn't really much better No OS i'm aware of has proper handling for notifying users and their applications of WHY their unmounts are failing.
And I found out the hard way that lsof won't tell you if the kernel itself has the file open. I spent ages trying to figure out why I couldn't unmount a partition before finally realising it was because I had an iso image on the partition loopback mounted.
what is your take with this device making up a beowulf cluster?
If you want to learn about running a cluster and go for it.
If you want to get computing done just buy a bloody i5 2500 and stick it on a cheap H61 mobo. The i5 has 4 cores at over four times the clockspeeds. So assuming the two architectures give similar performance per clock (i'd expect sandy bridge is faster but I dunno for sure) the i5 should be equivilent to over 16 pis. Further the Pi is limited by a USB based network connection.
I'd expect the real cost of a Pi model B to be over £30 once you add in VAT, shipping, SD card, network cable and USB cable (to cut up and connect to your PSU).
You can get an i5 2500 with mobo, 8GB ram and 8GB SSD for under £300 including VAT and delivery and as above i'd expect it to thrash a cluster of 10 Pis.
You will need a PSU either way. You may or may not want some form of case. If you do I'd expect a case for a standard PC would work out cheaper than a case for a cluser of Pis. You may or may not need additional storage (on top of the boot drives I included above) either way.
Unfortunately the Pi has very little GPIO. It's one of the compromises they made to keep the cost down (AIUI the chip has lots more GPIO but bringing it out would have required more layers). You may be able to hook up some specific shields that don't require much IO but afaict the only way you are going to make a generic "pi to ardunio shield adaptor" is to include either IO expanders or a microcontroller in it.
The way the EU vat system works is that as a VAT registered buisness you charge VAT on what you sell but you claim back VAT on what you buy. So the total VAT charged doesn't depend on the number of companies the product past through. There are slight complications around B2B sales between different countries in the EU but the basic principal holds.
This particular story though is not about VAT it's about import duty which depends on what customs category the goods fall into when they pass through customs into the EU. Performing manufacturing before importing will change the customs category and hence change the ammount of customs duty payable.
When you import something from outside the EU you pay both import duty and VAT (and VAT on the customs duty, and usually a handling charge to the carrier who cleared the package through customs). If you are a VAT registered buisness you claim the import VAT back and charge VAT on what you sell. Import duty however can't be claimed back under most circumstances (IIRC there are a few situations arround re-export where you can but I don't know the details).
Import duty varies depending on both the type of goods in question and country of origin with a huge number of confusing codes for different types of goods. Thankfully i've never imported enough stuff myself to have to deal with it.
it's probably more geared toward people that want to geek out but can't afford to.
Computers in schools have been taken over by IT departments and many parents would not want their kids "geeking out" on their main computer. Most kids can't afford to buy their own regular computers even in first world countries. Furthermore regular computers do not come set up to encourage programming. Programming environments are an optional extra (admittedly often a free one now but still you have to find and install them) and modern PCs make interfacing with your own hardware a PITA on both a hardware (paralell ports are fast dissapearing) and software (the NT line doesn't like you doing low level port access, there are hacks but....) perspective.
well $57 from a supplier who has no rating listed on the google product search, that i've never heard of an whose website looks rather dodgy. At best i'd say that is surpless stock at worst i'd say it's an outright scam. All the other sellers listed are FAR more expensive.
The big difference between Ebert and hawkings situation is that ebert was only forced to use a speech synth fairly recently. So probablly more people are going to know him by his natural voice than by the synth he uses.
Hawking has been using that speech synth for many years and I strongly suspect more people know him by that voice than know him by his natural voice.
AMD and intel have the same instruction set architectures. Yeah the internal microarchitecture is different but then that is also true of different generations from the same vendor and it doesm't really matter for the user.
jumped at the concept of switching architectures on the suggestion of an industry shill at a conference?
It sounds like intel offered to give him a new* and better computer do the work of upgrading him to said better computer (which is probablly quite considerable given the system built for him arround it) and then support that computer afterwards for free. I'm not surprised he took the offer.
*Afaict laptops that are carted arround all the time have a finite lifetime so in an application like this you probablly do want to replace the laptop every so often.
Decent sockets may cost more than the chip you're socketing.
For very cheap chips the socket may cost less than the chip but at least the sockets I always use aren't all that expensive and are usually a small part of the price of a prototype PCB assembly.
The only worthwhile, decently performing sockets, with low parasitics, are the individual turned pin sockets that you solder one-by-one directly into the PC board holes. Those are not cheap.
Maybe so but this board probablly doesn't need low parasitics. All those connectors etc are probablly going to have far more parasitics than a DIL socket anyway.
Anything else will markedly decrease reliability.
Can't say i've ever had problems with ordinary turned pin sockets bought from my regular electronics suppliers. Don't think i've ever had a problem with the cheap sockets that have sprung contacts made from bent metal either.
For a devboard i'll take the ease of troubleshooting and modification provided by socketed dil chips over some supposed reliability reduction that i've never noticed any day.
the 64-bit version isn't compatible with a lot of 32-bit software and generally causes headaches.
Out of interest when did you last try it and what problems did you have? Afaict most of the work done to add support for 64-bit vista and win7 also benefitted XP proffessional x64 edition (which is technically a version of 2K3 not XP).
And yes I do run it on my office computer most of the time (I also have a 32-bit XP install on there), I didn't fancy putting win7 on my main machine before i'd tried it on any other machines (I have tried it on other machines since and there are some things I like about it and other things I hate about it) but I needed more usable ram.
So far since running it i've run into just a handfull of scenarios where I had problems that according to the vendors I wouldn't have had on 64-bit win7. The first was with a data translation DT9816 that worked fine with the low level APIs but wouldn't work with the higher level APIs. Another was the NI mydaq which just wouldn't install at all. MPLAB pops up "MPLAB IDE may not support this OS" on startup but it seems to work fine with my ICD3 (using the drivers intended for 64-bit vista/win7).
Actually, both UNIX and Windows have two time formats:
the one that is more-or-less a count of second-or-sub-second intervals since a fixed time in the past;
the one that's year/month/day/hour/minute/second/maybe fractions of a second.
They do but my understanding was that the former was the internal format while the latter was mostly for display.
Also another BIG advantage of DIL is that it can be socketed. That means it's much easier to replace a part if you fry it. It also means if a board doesn't work you can pull chips out either to let you test the board itself for shorts etc or to allow you to test sections of the circuit in isolation.
I'll use surface mount if i'm space constrained or can't get the part I need in a DIL package but I much preffer DIL.
"Winging a pin" (removing the pin from the board and connecting to a wire) to change the circuit is also much easier on a socketed dil than on a SOIC.
Soldering a 1.27mm pitch SMT component is really easy, it takes about the same amount of time as a DIP component, and is much, much, smaller.
The thing I find takes the time with SOIC components is getting them aligned and keeping them aligned while I do the first couple of pins (diagonally opposite of coutse). Once you've done that first pin the rest is pretty easy but i've always found the first pin a PITA.
I'm not sure if the client for the upgraded DB that will run on Vista or win7, will run on XP
I'd be somewhat surprised if a buissness software vendor wasn't supporting XP at this point but stranger things have happened.
We're in a can't afford to upgrade and can't afford not to situation.
A few other options that you may or may not have considered.
Firstly dell will still sell you computers with XP drivers available. Afaict they can't sell off the shelf pre-downgaded machines anymore but you can still downgrade them yourself and if you have an existing dell XP pro CD there will be no activation issues. http://www.microsoft.com/oem/en/licensing/sblicensing/pages/downgrade_rights.aspx
MS also offer "XP mode" which is an XP vm running under virtual PC and is available for 7 pro/enterprise/ultimate users at no extra charge. Have you tried running the client for your database under this.
And of course there is the option of just replacing the bloody caps.
IMO there are a couple of major problems with that idea.
Firstly for an OS the part of "support" that most people care about is security updates. If a non-security bug hasn't been found in the 8 years after a program has been released then it's probablly a bug that most people can live with but running an OS without security patches in a networked environment is often a rather risky proposition.
How do you think it would go down if MS said to the general public "sure we have an XP patch for the security hole used by the blaster reloaded virus but to get it you will have to buy a superextended support contract for £200/machine/year?" I'm guessing it would go down even worse than saying "sorry we don't support XP anymore".
Secondly if a buisness is expected to pay £200 per year for an XP machine with support/security patches they are going to try and minimise the number of them. So while MS may initially get quite a bit of money from doing that I expect it would quickly dwindle.
If they can pinpoint and track you through the stores over their microcells, bluetooth or maybe even WLAN (if available)
I suspect that tracking customers through a shopping center using their mobile phones would be pretty crude. If the shopping center was one big anechoic chamber then radio signals could be triangulated but in practice I would expect multipath to mean that the best you can do is to figure out which receiver they are probablly closest to and maybe get a very rough idea of their distance from it. I highly doubt they could distinguish who is at which till in a busy store with any reasonable degree of reliability.
And even if they could I suspect that it would be difficult to get all the retailers to turn over their credit card retailers to the shopping centre.
Afaict both windows and unix use core time formats that assume 86400 seconds in a day and simply cannot represent leap seconds. So even if leap second support is present in the OS it will be papered on and I wouldn't want to rely on it without extensive testing:(
Some OSes don't support leap seconds, which complicates matters.
Both unix and windows use time formats that assume there are 86400 seconds in a day. There may be papered on support for leap seconds but I wouldn't want to rely on it without extensive testing of the application/OS combination in question:(.
Slashdot Mirror
Other projects, as requested: Bifferboard [bifferos.co.uk] and Beagleboard [beagleboard.org]. Both longer-established and actually in production. Bifferboard costs the same as Raspberry Pi Model B.
Essentially with the Pi you are getting a device with specs comparable to the beableboard but a price comparable to (probablly lower than but we won't know for sure until we see final shipped prices from the Pi guys) the bifferboard.
Agreed
According to the actual proposal they plan to support them for 54 weeks with an overlap of 12 weeks.
That is very little overlap between releases in which to plan your upgrade policy and since the changes won't come at any particular time of year it will be difficult to tie it in with other upgrades.
The problem comes when the import duty for the constituent parts/materials works out more expensive than the import duty for what is being manufactured. Then you have a disincentive to manufacture locally.
IMO a sane import duty system would have relatively high import duties for finished goods and lower import duties the further back along the production chain you moved so the incentive was to do as much of the manufacturing as possible in-country but afaict duties are set to pander to specific interests so they often don't end up sane.
Another problem is that many countries have exemptions for small imports. Even if they don't it's much easier to get away with dodgy declarations on small imports. This make sense from the point of view of saving paperwork (often the charge for collecting import duty/VAT is more than the duty/VAT itself) but it also creates an incentive for customers to buy directly from abroad.
If you don't know what your computer is doing and why it would be writing to a removable device then that's your own failing.
I'm not the original poster and i've never had issues with removable media that I couldn't find an application to blame for (some apps seem to keep files open even after the user has "closed" the visible representation of the file in the program). I'd bet very few users really know what is running in the background on their computer and what it is doing. Windows is worse because apps tend to be closed source and not filtered by a distro but even on linux do you really know the exact details of how every program you have installed behaves?
And afaict a file open for read is more than enough to prevent unmouning a partition on both windows and linux. It doesn't need to be writing it.
Linux does that just the same too - ever seen 'Device or resource busy'? No help there. Allowing you to force unmount is not a safe thing to do for casual users - you need to know what you're doing and be prepared to accept the consequences of a mistake.
If the OS lets the user force unmount then at least the FS gets the opertunity to shut down cleanly even if the app doesn't. If the OS doesn't let the user force unmount and the user pulls the plug anyway then neither the FS or the app gets to shut down cleanly.
Though as you say linux isn't really much better No OS i'm aware of has proper handling for notifying users and their applications of WHY their unmounts are failing.
And I found out the hard way that lsof won't tell you if the kernel itself has the file open. I spent ages trying to figure out why I couldn't unmount a partition before finally realising it was because I had an iso image on the partition loopback mounted.
what is your take with this device making up a beowulf cluster?
If you want to learn about running a cluster and go for it.
If you want to get computing done just buy a bloody i5 2500 and stick it on a cheap H61 mobo. The i5 has 4 cores at over four times the clockspeeds. So assuming the two architectures give similar performance per clock (i'd expect sandy bridge is faster but I dunno for sure) the i5 should be equivilent to over 16 pis. Further the Pi is limited by a USB based network connection.
I'd expect the real cost of a Pi model B to be over £30 once you add in VAT, shipping, SD card, network cable and USB cable (to cut up and connect to your PSU).
You can get an i5 2500 with mobo, 8GB ram and 8GB SSD for under £300 including VAT and delivery and as above i'd expect it to thrash a cluster of 10 Pis.
You will need a PSU either way. You may or may not want some form of case. If you do I'd expect a case for a standard PC would work out cheaper than a case for a cluser of Pis. You may or may not need additional storage (on top of the boot drives I included above) either way.
Unfortunately the Pi has very little GPIO. It's one of the compromises they made to keep the cost down (AIUI the chip has lots more GPIO but bringing it out would have required more layers). You may be able to hook up some specific shields that don't require much IO but afaict the only way you are going to make a generic "pi to ardunio shield adaptor" is to include either IO expanders or a microcontroller in it.
The way the EU vat system works is that as a VAT registered buisness you charge VAT on what you sell but you claim back VAT on what you buy. So the total VAT charged doesn't depend on the number of companies the product past through. There are slight complications around B2B sales between different countries in the EU but the basic principal holds.
This particular story though is not about VAT it's about import duty which depends on what customs category the goods fall into when they pass through customs into the EU. Performing manufacturing before importing will change the customs category and hence change the ammount of customs duty payable.
Import duty != VAT.
When you import something from outside the EU you pay both import duty and VAT (and VAT on the customs duty, and usually a handling charge to the carrier who cleared the package through customs). If you are a VAT registered buisness you claim the import VAT back and charge VAT on what you sell. Import duty however can't be claimed back under most circumstances (IIRC there are a few situations arround re-export where you can but I don't know the details).
Import duty varies depending on both the type of goods in question and country of origin with a huge number of confusing codes for different types of goods. Thankfully i've never imported enough stuff myself to have to deal with it.
it's probably more geared toward people that want to geek out but can't afford to.
Computers in schools have been taken over by IT departments and many parents would not want their kids "geeking out" on their main computer. Most kids can't afford to buy their own regular computers even in first world countries. Furthermore regular computers do not come set up to encourage programming. Programming environments are an optional extra (admittedly often a free one now but still you have to find and install them) and modern PCs make interfacing with your own hardware a PITA on both a hardware (paralell ports are fast dissapearing) and software (the NT line doesn't like you doing low level port access, there are hacks but....) perspective.
well $57 from a supplier who has no rating listed on the google product search, that i've never heard of an whose website looks rather dodgy. At best i'd say that is surpless stock at worst i'd say it's an outright scam. All the other sellers listed are FAR more expensive.
Customs rules are harmonised across the EU so if what the pi guys say is true for the UK it's probably true for the EU as a whole.
The big difference between Ebert and hawkings situation is that ebert was only forced to use a speech synth fairly recently. So probablly more people are going to know him by his natural voice than by the synth he uses.
Hawking has been using that speech synth for many years and I strongly suspect more people know him by that voice than know him by his natural voice.
the concept of switching architectures
AMD and intel have the same instruction set architectures. Yeah the internal microarchitecture is different but then that is also true of different generations from the same vendor and it doesm't really matter for the user.
jumped at the concept of switching architectures on the suggestion of an industry shill at a conference?
It sounds like intel offered to give him a new* and better computer do the work of upgrading him to said better computer (which is probablly quite considerable given the system built for him arround it) and then support that computer afterwards for free. I'm not surprised he took the offer.
*Afaict laptops that are carted arround all the time have a finite lifetime so in an application like this you probablly do want to replace the laptop every so often.
Decent sockets may cost more than the chip you're socketing.
For very cheap chips the socket may cost less than the chip but at least the sockets I always use aren't all that expensive and are usually a small part of the price of a prototype PCB assembly.
The only worthwhile, decently performing sockets, with low parasitics, are the individual turned pin sockets that you solder one-by-one directly into the PC board holes. Those are not cheap.
Maybe so but this board probablly doesn't need low parasitics. All those connectors etc are probablly going to have far more parasitics than a DIL socket anyway.
Anything else will markedly decrease reliability.
Can't say i've ever had problems with ordinary turned pin sockets bought from my regular electronics suppliers. Don't think i've ever had a problem with the cheap sockets that have sprung contacts made from bent metal either.
For a devboard i'll take the ease of troubleshooting and modification provided by socketed dil chips over some supposed reliability reduction that i've never noticed any day.
nice tip
the 64-bit version isn't compatible with a lot of 32-bit software and generally causes headaches.
Out of interest when did you last try it and what problems did you have? Afaict most of the work done to add support for 64-bit vista and win7 also benefitted XP proffessional x64 edition (which is technically a version of 2K3 not XP).
And yes I do run it on my office computer most of the time (I also have a 32-bit XP install on there), I didn't fancy putting win7 on my main machine before i'd tried it on any other machines (I have tried it on other machines since and there are some things I like about it and other things I hate about it) but I needed more usable ram.
So far since running it i've run into just a handfull of scenarios where I had problems that according to the vendors I wouldn't have had on 64-bit win7. The first was with a data translation DT9816 that worked fine with the low level APIs but wouldn't work with the higher level APIs. Another was the NI mydaq which just wouldn't install at all. MPLAB pops up "MPLAB IDE may not support this OS" on startup but it seems to work fine with my ICD3 (using the drivers intended for 64-bit vista/win7).
Actually, both UNIX and Windows have two time formats:
the one that is more-or-less a count of second-or-sub-second intervals since a fixed time in the past;
the one that's year/month/day/hour/minute/second/maybe fractions of a second.
They do but my understanding was that the former was the internal format while the latter was mostly for display.
Also another BIG advantage of DIL is that it can be socketed. That means it's much easier to replace a part if you fry it. It also means if a board doesn't work you can pull chips out either to let you test the board itself for shorts etc or to allow you to test sections of the circuit in isolation.
I'll use surface mount if i'm space constrained or can't get the part I need in a DIL package but I much preffer DIL.
"Winging a pin" (removing the pin from the board and connecting to a wire) to change the circuit is also much easier on a socketed dil than on a SOIC.
Soldering a 1.27mm pitch SMT component is really easy, it takes about the same amount of time as a DIP component, and is much, much, smaller.
The thing I find takes the time with SOIC components is getting them aligned and keeping them aligned while I do the first couple of pins (diagonally opposite of coutse). Once you've done that first pin the rest is pretty easy but i've always found the first pin a PITA.
I'm not sure if the client for the upgraded DB that will run on Vista or win7, will run on XP
I'd be somewhat surprised if a buissness software vendor wasn't supporting XP at this point but stranger things have happened.
We're in a can't afford to upgrade and can't afford not to situation.
A few other options that you may or may not have considered.
Firstly dell will still sell you computers with XP drivers available. Afaict they can't sell off the shelf pre-downgaded machines anymore but you can still downgrade them yourself and if you have an existing dell XP pro CD there will be no activation issues. http://www.microsoft.com/oem/en/licensing/sblicensing/pages/downgrade_rights.aspx
MS also offer "XP mode" which is an XP vm running under virtual PC and is available for 7 pro/enterprise/ultimate users at no extra charge. Have you tried running the client for your database under this.
And of course there is the option of just replacing the bloody caps.
IMO there are a couple of major problems with that idea.
Firstly for an OS the part of "support" that most people care about is security updates. If a non-security bug hasn't been found in the 8 years after a program has been released then it's probablly a bug that most people can live with but running an OS without security patches in a networked environment is often a rather risky proposition.
How do you think it would go down if MS said to the general public "sure we have an XP patch for the security hole used by the blaster reloaded virus but to get it you will have to buy a superextended support contract for £200/machine/year?" I'm guessing it would go down even worse than saying "sorry we don't support XP anymore".
Secondly if a buisness is expected to pay £200 per year for an XP machine with support/security patches they are going to try and minimise the number of them. So while MS may initially get quite a bit of money from doing that I expect it would quickly dwindle.
If they can pinpoint and track you through the stores over their microcells, bluetooth or maybe even WLAN (if available)
I suspect that tracking customers through a shopping center using their mobile phones would be pretty crude. If the shopping center was one big anechoic chamber then radio signals could be triangulated but in practice I would expect multipath to mean that the best you can do is to figure out which receiver they are probablly closest to and maybe get a very rough idea of their distance from it. I highly doubt they could distinguish who is at which till in a busy store with any reasonable degree of reliability.
And even if they could I suspect that it would be difficult to get all the retailers to turn over their credit card retailers to the shopping centre.
Cause that's something you do so often, right.
When trying to browse the web on a small screen it certainly is. Zoom out to see the whole page, zoom in to actually read the damn thing.
Afaict both windows and unix use core time formats that assume 86400 seconds in a day and simply cannot represent leap seconds. So even if leap second support is present in the OS it will be papered on and I wouldn't want to rely on it without extensive testing :(
Some OSes don't support leap seconds, which complicates matters.
Both unix and windows use time formats that assume there are 86400 seconds in a day. There may be papered on support for leap seconds but I wouldn't want to rely on it without extensive testing of the application/OS combination in question :(.