The most likely scenario for a damaging asteroid impact is a city killer. A dangerous asteroid impact like the Tunguska event has about as much power as a single nuclear bomb. Which is enough to take out an entire metropolitan area such an asteroid lands in the middle of a metropolitan area. If we know where an asteroid will hit in advance we can evacuate cities. Months or years of lead time is not hard so we should concentrate on looking for these things.
Why does the conversation always talk of deflection? And making us feel scared because our space program is not up to snuff? When all we need to do is hop in a car and drive away?
In the general case it is true that you can not tell if an program will halt. However most programs do halt and most programmers understand why. Theorem provers present a language to their users that does not admit the general case in which halting is guaranteed which is part of a more important guarantee that all programs can be reduced to a normal form.
Which means that the halting problem is solved for a program written in a language like coq.
The practical challenge seems to be that formal detailed requirements and specifications for algorithms are about as hard to write as the algorithms themselves.
A functional justice system keeps us from descending to the level of personal vengeance and feuding.
It is very sad to see the Justice system failing here.
No one has officially called these prosecutors out on their failings in any other way so we get this. I don't think harassment of these prosecutors and MIT and JSTOR is the appropriate reaction. Nor do I think it is the appropriate reaction that the prosecutors have not been reprimanded and appropriate taken to keep non-sense like this from happening.
How high does this failure in the Justice system go?
The statement that implied DNS servers can implement this is bunk. However BCP38 is the fix. The attack would have been impossible without spoofed IP source addresses.
An application/reflection denial of service attack is actually possible with SNMP and several other protocols. Even if all of the DNS servers were closed this attack could happen.
There is no extra fee to use wifi calling on T-mobile.
So if you have a wifi network in your house you have already setup a call extender that can be used with T-mobile. It can be particularly nice if you travel outside of the US and want to call home. You get local US domestic rates.
Manually assigned ipv6 is quite doable. It is really just a matter of assigning $PREFIX::$small_memorable_number.
There should only be one prefix you have to worry about and if you forget it you can look at any other computer on the network. Then just assign your servers each a small number.
For your case with VMs coming and going it would not be at all hard (and would probably result in better testing) to go the ISP route and assign a unique name to every address and then just report that name to your testers and devs. Reusing the name is exactly the same as reusing the ip address. Then you just have a series of machine names. testvm1, testvm2, testvm3,... etc.
Really none of this is very hard, confusing or cumbersome. It just takes someone asking: "How do I make this work?" instead of thinking "Oh no! that is going to be horrible." and looking for excuses not to make it work.
Obama has a vision a space program that fits within the budget.
The Republicans have a vision. The space shuttle pork still flowing despite the cancellation of the space shuttle. The republicans call the lack of pork lack of vision because the can't see anything to eat. The Republicans want a return to Apollo where the pork flowed more freely.
Now if someone really wants vision let me propose this. Charge NASA with laying the groundwork for colonizing the solar system. This should include the research (aka robotic probes) to figure out what is out there. This should include making space flight affordable and accessible without being a member of NASA.
Fundamentally space flight is affordable. Right now the fuel cost for a single person to orbit is about $70,000.
One way trips to Mars can be made as cheap as $500,000 if you believe Elon Musk.
A trip to Mars reduce to $500,000 is accessible to the middle class in the United States. Accessible to 100 million or more people. A colonization trip to Mars for $500,000 starts sounding like a good deal on a house given how silly housing prices are on the east and west coast of the United States.
So let's make the vision space flight and space colonization for the middle class. Let's laugh at everyone who suggests the vision for NASA is to give hand outs to the incumbent space companies and their over priced rockets.
Somehow the analogy seems apt. Tunes is a decade old and still has not gotten anywhere. Although the tunes survey subproject has at times been interesting, and was a good resource before wikipedia came along.
With a little luck the EDOS project will be more grounded and a little more down to earh.
Hmm. The more I think about it this looks like a funding hack by mandrake to get other organizations to help them build and test their distribution. Most of the things they were complaining about did not sound fundamental to open source development but did sound a lot like problems a distribution vendor would have.
I guess time will tell if this is a cool practical hack that supports mandrake. Or some academic proof of concept project that is generally useless for getting things done.
The current opteron memory bandwidth per memory
controller is 6400MB/s.
So it is twice that in a 2 way system, and four
times that in a 4 way system. At least if properly configured.
The memory bandwidth in an Itanium2 system with it's shared FSB is 6400MB/s total.
Opteron should do much better in 2-way and 4-way benchmarks once OS's begin to optimize for a NUMA
architecture. But even with out that the performance is quite good.
The only thing I have seen that the Itanium2 has is a larger physical address space support. Which is great if you are building a 512 CPU box and irrelevant everywhere else.
I think OF is a fine bootloader, and I have no problems having it ride on top of LinuxBIOS.
But there is a problem of dynamics.
The problem. There is limited time to develop a BIOS for a new board. So firmware developers have a limited amount of time to develop something. People don't really care about the
bios because it is just an intermediate step
to get your kernel booted.
So the way to make it reliable is to make things
as simple as possible.
The sheer unreliability of the current crop of BIOS's is the issue that got LinuxBIOS moving. A typical failure rate is about 1 in 100 boots fail. And no amount of clean interfaces or good design will let me fix problems that come up because of buggy implemenations or buggy hardware. But give me the code and I can fix it.
The interface and the features that come after the hardware is in a usable state. That is a minor matter as long as it is well enough factored that it is board indpendent.
As far as interfaces I totally agree that LinuxBIOS should not be ``the'' standard but simply one implementation of it. But things don't change unless someone attempts it. I won't propose LinuxBIOS as a general purpose BIOS replacement until it can support it. But things are slowly moving in the direction of making that a real possibility.
If you want Openfirmware don't order on a pc don't order a board without it.
As for support LinuxBIOS is supported. And I suspect I have better connections at AMD and Intel than most of your BIOS developers.
As for bootloaders don't forget that in their limited ways OF, EFI, and the PC-BIOS are all OS's. And if I have to use an OS to boot I would much rather use a real one.
LinuxBIOS is quite flexible. The difference though it that it is well factored. The core
LinuxBIOS piece turns on RAM, initializes the hardware, exports tables of motherboard specific information and in general makes the machine useable. The the core of LinuxBIOS jumps to
a bootloader.
And that bootloader can be anything you want. Etherboot and the kernel are the prefered bootloaders right now. But we have bootloaders in progress that implement OpenFirmware, and that provide a legacy pc interface.
LinuxBIOS was not designed as a means to an end.
LinuxBIOS is used as a means to an end, so we start small with the simplest implemenation we can. And then as more time and resources become
available we make it more interesting.
I find it funny that you argue for the bloat features of open firmware on the one hand and then argue against bloat on the other.
For 64bit PC's doing firmware projects is actually easier. AMD has integrated most of the hard components to support into the processor itself so the code only has to be written once. And AMD is open source friendly, and of course LinuxBIOS already supports the k8.
Most of the lag of getting a platform supported comes from the intercompany politics that is needed to get the documentation. This is happening slowly but surely.
Beyond that testing on clusters has one interesting effect. What BIOS manufacturer can afford a multi-million dollar test platform to make certain their BIOS reliably boots?
Eric
Actually I'm willing to guess it hasn't done
more because simply because it has a new, design
and it takes people a while to get used to things
like this.
From what I can tell uptake has been snowballing for a long time. The snowball just started very very small...:)
As for booting windows it could if anyone was sufficiently motivated but I that doesn't appear
to be the niche of LinuxBIOS.
Nope I've done it several times on 960 nodes in MCR,it hasn't been a problem. In fact I can completely reinstall MCR (which being the conservative cousine of PINK has disk) in an hour or so. This includes all of the time for fighting problems.
Cost + Capability.
MCR the conservative cousin to pink will likely how up at #5 on the the top500 list at the next posting.
And at roughly 2000 CPUs it has the fewer CPUs than the risk competition it is beeting, and/or is beaten by.
Slashdot Mirror
The most likely scenario for a damaging asteroid impact is a city killer. A dangerous asteroid impact like the Tunguska event has about as much power as a single nuclear bomb. Which is enough to take out an entire metropolitan area such an asteroid lands in the middle of a metropolitan area. If we know where an asteroid will hit in advance we can evacuate cities. Months or years of lead time is not hard so we should concentrate on looking for these things.
Why does the conversation always talk of deflection? And making us feel scared because our space program is not up to snuff? When all we need to do is hop in a car and drive away?
In the general case it is true that you can not tell if an program will halt. However most programs do halt and most programmers understand why. Theorem provers present a language to their users that does not admit the general case in which halting is guaranteed which is part of a more important guarantee that all programs can be reduced to a normal form.
Which means that the halting problem is solved for a program written in a language like coq.
The practical challenge seems to be that formal detailed requirements and specifications for algorithms are about as hard to write as the algorithms themselves.
A functional justice system keeps us from descending to the level of personal vengeance and feuding.
It is very sad to see the Justice system failing here.
No one has officially called these prosecutors out on their failings in any other way so we get this. I don't think harassment of these prosecutors and MIT and JSTOR is the appropriate reaction. Nor do I think it is the appropriate reaction that the prosecutors have not been reprimanded and appropriate taken to keep non-sense like this from happening.
How high does this failure in the Justice system go?
The statement that implied DNS servers can implement this is bunk. However BCP38 is the fix. The attack would have been impossible without spoofed IP source addresses.
An application/reflection denial of service attack is actually possible with SNMP and several other protocols. Even if all of the DNS servers were closed this attack could happen.
There is no extra fee to use wifi calling on T-mobile.
So if you have a wifi network in your house you have already setup a call extender that can be used with T-mobile. It can be particularly nice if you travel outside of the US and want to call home. You get local US domestic rates.
How terribly insensitive and how clueless can the whitehouse be?
There should only be one prefix you have to worry about and if you forget it you can look at any other computer on the network. Then just assign your servers each a small number.
For your case with VMs coming and going it would not be at all hard (and would probably result in better testing) to go the ISP route and assign a unique name to every address and then just report that name to your testers and devs. Reusing the name is exactly the same as reusing the ip address. Then you just have a series of machine names. testvm1, testvm2, testvm3, ... etc.
Really none of this is very hard, confusing or cumbersome. It just takes someone asking: "How do I make this work?" instead of thinking "Oh no! that is going to be horrible." and looking for excuses not to make it work.
The Republicans have a vision. The space shuttle pork still flowing despite the cancellation of the space shuttle. The republicans call the lack of pork lack of vision because the can't see anything to eat. The Republicans want a return to Apollo where the pork flowed more freely.
Now if someone really wants vision let me propose this. Charge NASA with laying the groundwork for colonizing the solar system. This should include the research (aka robotic probes) to figure out what is out there. This should include making space flight affordable and accessible without being a member of NASA.
Fundamentally space flight is affordable. Right now the fuel cost for a single person to orbit is about $70,000.
One way trips to Mars can be made as cheap as $500,000 if you believe Elon Musk.
A trip to Mars reduce to $500,000 is accessible to the middle class in the United States. Accessible to 100 million or more people. A colonization trip to Mars for $500,000 starts sounding like a good deal on a house given how silly housing prices are on the east and west coast of the United States.
So let's make the vision space flight and space colonization for the middle class. Let's laugh at everyone who suggests the vision for NASA is to give hand outs to the incumbent space companies and their over priced rockets.
Somehow the analogy seems apt. Tunes is a decade old and still has not gotten anywhere. Although the tunes survey subproject has at times been interesting, and was a good resource before wikipedia came along.
With a little luck the EDOS project will be more grounded and a little more down to earh.
Hmm. The more I think about it this looks like a funding hack by mandrake to get other organizations to help them build and test their distribution. Most of the things they were complaining about did not sound fundamental to open source development but did sound a lot like problems a distribution vendor would have.
I guess time will tell if this is a cool practical hack that supports mandrake. Or some academic proof of concept project that is generally useless for getting things done.
The current opteron memory bandwidth per memory controller is 6400MB/s. So it is twice that in a 2 way system, and four times that in a 4 way system. At least if properly configured. The memory bandwidth in an Itanium2 system with it's shared FSB is 6400MB/s total. Opteron should do much better in 2-way and 4-way benchmarks once OS's begin to optimize for a NUMA architecture. But even with out that the performance is quite good. The only thing I have seen that the Itanium2 has is a larger physical address space support. Which is great if you are building a 512 CPU box and irrelevant everywhere else.
I think OF is a fine bootloader, and I have no problems having it ride on top of LinuxBIOS. But there is a problem of dynamics. The problem. There is limited time to develop a BIOS for a new board. So firmware developers have a limited amount of time to develop something. People don't really care about the bios because it is just an intermediate step to get your kernel booted. So the way to make it reliable is to make things as simple as possible. The sheer unreliability of the current crop of BIOS's is the issue that got LinuxBIOS moving. A typical failure rate is about 1 in 100 boots fail. And no amount of clean interfaces or good design will let me fix problems that come up because of buggy implemenations or buggy hardware. But give me the code and I can fix it. The interface and the features that come after the hardware is in a usable state. That is a minor matter as long as it is well enough factored that it is board indpendent. As far as interfaces I totally agree that LinuxBIOS should not be ``the'' standard but simply one implementation of it. But things don't change unless someone attempts it. I won't propose LinuxBIOS as a general purpose BIOS replacement until it can support it. But things are slowly moving in the direction of making that a real possibility. If you want Openfirmware don't order on a pc don't order a board without it. As for support LinuxBIOS is supported. And I suspect I have better connections at AMD and Intel than most of your BIOS developers. As for bootloaders don't forget that in their limited ways OF, EFI, and the PC-BIOS are all OS's. And if I have to use an OS to boot I would much rather use a real one.
LinuxBIOS is quite flexible. The difference though it that it is well factored. The core LinuxBIOS piece turns on RAM, initializes the hardware, exports tables of motherboard specific information and in general makes the machine useable. The the core of LinuxBIOS jumps to a bootloader. And that bootloader can be anything you want. Etherboot and the kernel are the prefered bootloaders right now. But we have bootloaders in progress that implement OpenFirmware, and that provide a legacy pc interface. LinuxBIOS was not designed as a means to an end. LinuxBIOS is used as a means to an end, so we start small with the simplest implemenation we can. And then as more time and resources become available we make it more interesting. I find it funny that you argue for the bloat features of open firmware on the one hand and then argue against bloat on the other. For 64bit PC's doing firmware projects is actually easier. AMD has integrated most of the hard components to support into the processor itself so the code only has to be written once. And AMD is open source friendly, and of course LinuxBIOS already supports the k8. Most of the lag of getting a platform supported comes from the intercompany politics that is needed to get the documentation. This is happening slowly but surely. Beyond that testing on clusters has one interesting effect. What BIOS manufacturer can afford a multi-million dollar test platform to make certain their BIOS reliably boots? Eric
Actually I'm willing to guess it hasn't done more because simply because it has a new, design and it takes people a while to get used to things like this. From what I can tell uptake has been snowballing for a long time. The snowball just started very very small... :)
As for booting windows it could if anyone was sufficiently motivated but I that doesn't appear
to be the niche of LinuxBIOS.
With LinuxBIOS the final kernel does not reside in flash, so you don't need to flash your BIOS to upgrade.
Which is one of the major points of pink to resolve the scalability issues.
Nope I've done it several times on 960 nodes in MCR,it hasn't been a problem. In fact I can completely reinstall MCR (which being the conservative cousine of PINK has disk) in an hour or so. This includes all of the time for fighting problems.
Cost + Capability. MCR the conservative cousin to pink will likely how up at #5 on the the top500 list at the next posting. And at roughly 2000 CPUs it has the fewer CPUs than the risk competition it is beeting, and/or is beaten by.