Who did not already understand that floats are approximations?
Floats are not, in general, approximations. Any float of the form a*2^(-1) + b*2^(-2) + c*2^(-3), etc. where a, b, c, etc. are 0 or 1, up to the limits of number of bits in the fraction component of the word, is an exact representation of the number. Operations on floats can yield exact results as well.
I think your problem is that you have no concept of how real-time systems and robots are built. Most of them, sooner or later, become collections of dozens of embedded systems, and many of those need to be programmed individually.
There is no "sooner or later" in a one-day competition. Like any project, you look at the requirements and then you choose implementation, not the other way around. If you proposed to your team that this vehicle should be built using "dozens of embedded systems", I'd be surprised if anyone agreed.
> Got any pointers to this large body of robotics code that only runs on Linux? This is a good start to find out about real-time and embedded Linux: http://www.linuxdevices.com/
I'm familiar with that site. But I didn't find a large body of Linux-only robotics code there. Got a more specific link?
Yes, and the OS kernel needs to be modified to run, say, on one of the many embedded systems that you can buy.
Many dev kits are sold with one or more OSes supported by the manufacturer. The one used by the winning team apparently had a version of XP embedded available.
For example, we're using a bunch of intelligent cameras in our work, and they run embedded Linux inside the camera. We have dozens of other embedded systems running Linux that couldn't possibly run Windows. And those embedded systems don't run stripped down Linux systems, they are fully functional, comfortable Linux environments.
Sounds like a cool job.
Even if you managed to put Windows on one of those devices, it would, for all practical purposes, still be a different OS from anything you are used to.
All I care is that it fits my requirements. I don't care if it isn't the same as a standard Windows desktop machine. Heck, I'd hope that they'd cut out all the bloat.
The issues are licensing, source code accessibility, range of available hardware platforms that the software runs on, minimum hardware requirements, and user community.
"Does the solution solve the problem?" That's the question that matters to me in the end.
In this case, licensing is end-user only: they aren't selling the thing. Clearly, the winning team in this contest found a Windows-based SBC that suited their requirements, which covers the source code/hardware platforms/hardware requirements issues. Having a "user community" is useful, but unless you are copy-and-paste hackers, most coding/technical issues are going to be solved in-house.
I'm certain the winning team here could have also implemented their solution using Linux. For whatever reason (lack of Linux knowledge, free hardware/software donations, huge monetary bribes directly from Bill Gates?), they chose not to.
I certainly believe you when you say you couldn't possibly implement your work with Windows devices. That's not what we are talking about though.
And, I'm sorry, if you think that those issues don't matter when building or delivering products, you are both inexperienced and naive.
I realize I may not have as much experience as you. 20 years of development isn't very much, I know. And having worked on videogames that sold millions of copies, I probably don't have the extensive product development experience you do. But try to keep the personal insults out of an otherwise interesting discussion.
People may or may not need to modify the OS themselves. But what they definitely need is an OS that runs on a variety of different embedded devices, chipsets, etc.
When you have an autonomous vehicle project to complete, you need an OS that runs on exactly ONE piece of hardware, not a variety. If a particular platform (hardware + OS) suits your requirements, it should obviously be considered.
They also need an OS that there is a large existing body of code for real-time robotics tasks.
How are robotics tasks related to the OS? Got any pointers to this large body of robotics code that only runs on Linux?
The fact that Linux is open source has allowed vendors and research groups to create and share the code and modifications that address those needs.
Only in cases where the OS itself is modified does access to the OS source code matter. Are you talking about application-level code or kernel code?
In any case, you're asking the wrong question. The question shouldn't be "why Linux", the question should be "why Windows". It's just a fact that Linux is dominating this space, and the fact that it's free is an added bonus. I simply don't see any reason to use Windows; what advantages do you think Windows has over Linux for these applications?
I wasn't asking "Why Linux?" My point was that the choice of OS isn't all that important. It's been my experience coding on many platforms and OSes over the last two decades that the costs, development time, and ultimate success of a project has little to do with the OS and lots to do with the people on the team.
That has nothing to do with technological differences, it has to do with community, source code, non-disclosure agreements, and tools, and those are quite limited for Windows XP compared to Linux. Using Windows XP to run an autonomous vehicle is quite unusual because a lot of non-technological constraints make it a hard thing to do.
You lost me. First, I have no idea why a team needs a "community" to develop an autonomous vehicle. I assume your references to "source code" and "non-disclosure agreements" implies that one would need to modify the OS in order to implement an autonomous vehicle, which is extremely unlikely. Even if you needed to write a driver for specialized hardware, Windows has a DDK which is well documented, with sample source code, all downloadable. And MS gives away a version of Visual Studio nowadays. So, what are the real barriers you are talking about?
By the way, I'm writing this from my Debian desktop which I use daily at work.
> Where have I heard this (http://en.wikipedia.org/wiki/Phantom_Game_Console ) before?
"I'd be amazed if the PlayStation 4 has a physical disc drive." - Sony's Phil Harrison. "I'd be amazed if the Phantom Console has a physical embodiment." - overheard.
I live just down the street from Criminal Records. It's right next to Savage Pizza. Really. It's a very cool independent shop. They know their music. Sadly, I don't think their market is going to be saved by a new physical format.
First paragraph: Blogger Jailed After Defying Court Orders
By JESSE McKINLEY Published: August 2, 2006
SAN FRANCISCO, Aug. 1 -- A freelance journalist and blogger was jailed on Tuesday after refusing to turn over video he took at an anticapitalist protest here last summer and after refusing to testify before a grand jury looking into accusations that crimes were committed at the protest.
Each of those countries has more than 1 million children. In order to fit their "one laptop per child" criteria, there will be a lottery. The winners get the laptops. The losers get to choose a method of execution.
I'm sorry to tell you, but the US is a democracy, not a meritocracy. Thus, what the people want goes, not what is "right" scientifically. It's good and bad that way.
Last year, total exports of ethanol from Brazil were about 700 million gallons. Despite the high tarriff, the US actually imported over 100 million gallons of that. In contrast, the US produced about 4 billion gallons. Total Brazilian exports don't look to increase any time soon, due to increased domestic demand. The scale simply isn't there.
And lastly, most ethanol is produced by corn in an extremely wasteful process that requires an enormous amount of energy which cuts in on its overall efficiency.
Worldwide, most ethanol is actually produced by sugarcane. The corn thing is a US-specific thing based largely on the economics of government subsidies. Per-acre ethanol yields for sugar cane (Brazil and India) and sugar beets (France) are reportedly double that of corn in the US.
In my experience, even the firmware can be upgraded wirelessly. I've done this on Linksys, Netgear, and DLink routers. What usually can't be done (although it's generally configurable) is administrative access through the WAN port.
I live in a fairly large city, and we have nothing like a subway or street cars. There is a bus system, but it only covers a fairly small portion of down town. I would have to walk over 20 miles to get to the nearest bus stop, and I live in the city, not in a suburb.
Perhaps you technically live within some city limits. But if you are 20 miles from a bus stop, you live in the suburbs. Probably of Houston.
Slashdot Mirror
As soon as you call someone "out of their depth", you've pretty much defined yourself.
Who did not already understand that floats are approximations?
Floats are not, in general, approximations. Any float of the form a*2^(-1) + b*2^(-2) + c*2^(-3), etc. where a, b, c, etc. are 0 or 1, up to the limits of number of bits in the fraction component of the word, is an exact representation of the number. Operations on floats can yield exact results as well.
Nice troll kid.
I think your problem is that you have no concept of how real-time systems and robots are built. Most of them, sooner or later, become collections of dozens of embedded systems, and many of those need to be programmed individually.
There is no "sooner or later" in a one-day competition. Like any project, you look at the requirements and then you choose implementation, not the other way around. If you proposed to your team that this vehicle should be built using "dozens of embedded systems", I'd be surprised if anyone agreed.
> Got any pointers to this large body of robotics code that only runs on Linux?
This is a good start to find out about real-time and embedded Linux: http://www.linuxdevices.com/
I'm familiar with that site. But I didn't find a large body of Linux-only robotics code there. Got a more specific link?
Yes, and the OS kernel needs to be modified to run, say, on one of the many embedded systems that you can buy.
Many dev kits are sold with one or more OSes supported by the manufacturer. The one used by the winning team apparently had a version of XP embedded available.
For example, we're using a bunch of intelligent cameras in our work, and they run embedded Linux inside the camera. We have dozens of other embedded systems running Linux that couldn't possibly run Windows. And those embedded systems don't run stripped down Linux systems, they are fully functional, comfortable Linux environments.
Sounds like a cool job.
Even if you managed to put Windows on one of those devices, it would, for all practical purposes, still be a different OS from anything you are used to.
All I care is that it fits my requirements. I don't care if it isn't the same as a standard Windows desktop machine. Heck, I'd hope that they'd cut out all the bloat.
The issues are licensing, source code accessibility, range of available hardware platforms that the software runs on, minimum hardware requirements, and user community.
"Does the solution solve the problem?" That's the question that matters to me in the end.
In this case, licensing is end-user only: they aren't selling the thing. Clearly, the winning team in this contest found a Windows-based SBC that suited their requirements, which covers the source code/hardware platforms/hardware requirements issues. Having a "user community" is useful, but unless you are copy-and-paste hackers, most coding/technical issues are going to be solved in-house.
I'm certain the winning team here could have also implemented their solution using Linux. For whatever reason (lack of Linux knowledge, free hardware/software donations, huge monetary bribes directly from Bill Gates?), they chose not to.
I certainly believe you when you say you couldn't possibly implement your work with Windows devices. That's not what we are talking about though.
And, I'm sorry, if you think that those issues don't matter when building or delivering products, you are both inexperienced and naive.
I realize I may not have as much experience as you. 20 years of development isn't very much, I know. And having worked on videogames that sold millions of copies, I probably don't have the extensive product development experience you do. But try to keep the personal insults out of an otherwise interesting discussion.
BS.
1300 miles/5 h = 260 miles/h (or 418 km/h).
Good luck even getting from Barcelona to Paris in under 9 hours.
People may or may not need to modify the OS themselves. But what they definitely need is an OS that runs on a variety of different embedded devices, chipsets, etc.
When you have an autonomous vehicle project to complete, you need an OS that runs on exactly ONE piece of hardware, not a variety. If a particular platform (hardware + OS) suits your requirements, it should obviously be considered.
They also need an OS that there is a large existing body of code for real-time robotics tasks.
How are robotics tasks related to the OS? Got any pointers to this large body of robotics code that only runs on Linux?
The fact that Linux is open source has allowed vendors and research groups to create and share the code and modifications that address those needs.
Only in cases where the OS itself is modified does access to the OS source code matter. Are you talking about application-level code or kernel code?
In any case, you're asking the wrong question. The question shouldn't be "why Linux", the question should be "why Windows". It's just a fact that Linux is dominating this space, and the fact that it's free is an added bonus. I simply don't see any reason to use Windows; what advantages do you think Windows has over Linux for these applications?
I wasn't asking "Why Linux?" My point was that the choice of OS isn't all that important. It's been my experience coding on many platforms and OSes over the last two decades that the costs, development time, and ultimate success of a project has little to do with the OS and lots to do with the people on the team.
That has nothing to do with technological differences, it has to do with community, source code, non-disclosure agreements, and tools, and those are quite limited for Windows XP compared to Linux. Using Windows XP to run an autonomous vehicle is quite unusual because a lot of non-technological constraints make it a hard thing to do.
You lost me. First, I have no idea why a team needs a "community" to develop an autonomous vehicle. I assume your references to "source code" and "non-disclosure agreements" implies that one would need to modify the OS in order to implement an autonomous vehicle, which is extremely unlikely. Even if you needed to write a driver for specialized hardware, Windows has a DDK which is well documented, with sample source code, all downloadable. And MS gives away a version of Visual Studio nowadays. So, what are the real barriers you are talking about?
By the way, I'm writing this from my Debian desktop which I use daily at work.
> Where have I heard this (http://en.wikipedia.org/wiki/Phantom_Game_Console ) before?
"I'd be amazed if the PlayStation 4 has a physical disc drive." - Sony's Phil Harrison.
"I'd be amazed if the Phantom Console has a physical embodiment." - overheard.
I live just down the street from Criminal Records. It's right next to Savage Pizza. Really. It's a very cool independent shop. They know their music. Sadly, I don't think their market is going to be saved by a new physical format.
Bloggers get special PMITA treatment.
m l
http://www.nytimes.com/2006/08/02/us/02protest.ht
First paragraph:
Blogger Jailed After Defying Court Orders
By JESSE McKINLEY Published: August 2, 2006
SAN FRANCISCO, Aug. 1 -- A freelance journalist and blogger was jailed on Tuesday after refusing to turn over video he took at an anticapitalist protest here last summer and after refusing to testify before a grand jury looking into accusations that crimes were committed at the protest.
>> Nuclear weaponry isn't quite enhancing my life
> Chemotherapy? Nuclear power plants?
ERROR: Message not received. Repeat transmission in bold caps and emphasis to increase likelyhood of uptake.
NUCLEAR WEAPONRY ISN'T QUITE ENHANCING MY LIFE
End transmission.
I understand your need to be contrarian. I'm sorry I bothered you by agreeing with you.
>> the US is a democracy
> You do know this isn't true, right?... The US is an indirect democracy
Yeah. Like I said.
Each of those countries has more than 1 million children. In order to fit their "one laptop per child" criteria, there will be a lottery. The winners get the laptops. The losers get to choose a method of execution.
I'm sorry to tell you, but the US is a democracy, not a meritocracy. Thus, what the people want goes, not what is "right" scientifically. It's good and bad that way.
http://dictionary.reference.com/browse/affecters
No entry found for affecters.
Last year, total exports of ethanol from Brazil were about 700 million gallons. Despite the high tarriff, the US actually imported over 100 million gallons of that. In contrast, the US produced about 4 billion gallons. Total Brazilian exports don't look to increase any time soon, due to increased domestic demand. The scale simply isn't there.
> Plus, the US I guarantee will not be the only country growing corn to contribute to the ethanol supplies.
If the US, with such abundant land resources, can't produce enough energy using corn for itself, where on Earth are we going to import it from?
> Also algae grows in water. So if it's landmass you'd need a 100 square mile tank to hold it.
.1 square miles.
Or, 1000 tanks, each
> 97% of the landmass of the US, spread out over all the airable land in the world, isnt nearly as extreme.
Sarcasm, right?
And lastly, most ethanol is produced by corn in an extremely wasteful process that requires an enormous amount of energy which cuts in on its overall efficiency.
Worldwide, most ethanol is actually produced by sugarcane. The corn thing is a US-specific thing based largely on the economics of government subsidies. Per-acre ethanol yields for sugar cane (Brazil and India) and sugar beets (France) are reportedly double that of corn in the US.
The largest object in the universe has to be an instance of some MFC class.
A better design would be to replace every 100th image or so with a randomly chosen one from google image search.
And every 100th sentence or so is run through a Swedish Chef borkifier filter.
In my experience, even the firmware can be upgraded wirelessly. I've done this on Linksys, Netgear, and DLink routers. What usually can't be done (although it's generally configurable) is administrative access through the WAN port.
I live in a fairly large city, and we have nothing like a subway or street cars. There is a bus system, but it only covers a fairly small portion of down town. I would have to walk over 20 miles to get to the nearest bus stop, and I live in the city, not in a suburb.
Perhaps you technically live within some city limits. But if you are 20 miles from a bus stop, you live in the suburbs. Probably of Houston.