You're quite correct, there are uses for it. Protocols that have sufficient inherent redundancy and error recovery are reasonable for this kind of broadcast. But I've run into a number of instances where some "clever" engineer said "just use multicast and we won't need so many ports and switches!", especially for data where missed packets was considered a failure and unacceptable. Then the job was left for the next few years of failures and debugging. By the time they'd finished laying proprietary, unstable, and unsupportable checksumming and packet indexing and creative mishandling of their networks, they'd spent far more on lost services and engineering time than if they'd simply run TCP from the start.
In many instances, it's more robust to scale the TCP: a cautiously built architecture without excessive bandwidth on individual channels can scale quite well, with very modest devices acting as proxies for the TCP traffic. But, sadly, that doesn't seem to be a very "exciting" technology, and isn't prey to the optimistic underbudgeting I've seen with multicast projects, and the tendency to assume that what works with 20% of the expected on only 2 hosts in the same rack running entirely hand-built test software will work with dozens or hundreds of machines scattered to multiple data centers with divergent platforms and services on them.
Similar to the expansion of the US "wild west", we're due for years of backfilling and territory arguments. Look ahead to the owners of/8 address ranges having them confiscated. (MIT, for example, hardly needs it: they should be NAT'ing all their internal traffic anyway to prevent "computer science majors" from pulling stupid stunts like the David LaMacchia case (http://www.informit.com/articles/article.aspx?p=169520).
NAT is notoriously lighter weight to support than IPv6, and helps provide some border control of undesirable services from inside your network. Replacing the router infrastructure and the configuration tools for stable, legacy systems to support IPv6 is expensive and the benefits of IPv6 are frankly underwhelming. It's exciting "auto-configuration" is, in most cases, a horrendously bad idea for public facing systems, and private systems don't need it. Useful security features, such as IPsec, were backported to IPv4. And the robust technical features of IPsec seem to be overwhelmed by the far easier to use client behavior of PPTP.
Multicast? Oh, dear. Do _not_ get me started on the flaws of multicast programmers decided that the lack of information about missed packets in multicast forcing them to rewrite TCP, badly, as an unstable software layer on top of multicast.
Other folks will mention their favorite virtualization toolkits. I'd like to mention that if the students expect to do serious work, such as rebuilding interesting packages to get their feet wet, they may well need more than 4 Gig of space. They may need swap space if their laptop or host machine is underpowered, and after you get done installing compilers and development libraries and Gnome or KDE, there's surprisingly little space left on a 4 Gig disk. Also, if they attempt to build RPM's in the chrooted "mock" or "mach" environments for stability of build environments, that's another chunk of space.
So if your budget can stand it, try to get 8 Gig USB sticks.
New York City and LA are interesting issues. One problem is the conflict between state and local authorities: the state expects state wide coverage, and high bandwidth providers have zero interest in building and maintaining high speed infrastructure to rural areas: it's quite expensive.
Another is that there is a piranha-like feeding frenzy going on for network bandwidth in such core urban areas: the companies are interfering with each other, both inside and outside the courts and lobbyist's offices, to _block_ progress by their competitors. And the arguments range from profound (what role should government have in using eminent domain to allow fiber runs or cell phone towers) to completely ridiculous. (How dare a provider tell me I can't run my own SMTP server! I'll send any email I want, how dare you call it spam, you're infringing on my freedom of speech!)
Fair enough. Sweden, rather than being _small_, is merely modest. Yes, it's a large country for Europe, but it's still a pretty modest size. I'd be curious how they fund their network infrastructure.
The Netherlands are _small_, less than 15,000 squarae miles, densely populated and highly developed. You have to be, in order to keep pushing back the ocean every year. (And good luck with that if global warming makes the oceans rise noticeably!) It is much easier to build dense, effective infrastructure to support such bandwidth in such a small area than it is in the USA, which has not only federal issues but 50 different state governments to negotiate with, and the states have vastly different requirements.
But cockroaches are not one species, any more than primates are all one species. They are an "order" a set of species with some common characteristics. Like sharks, they're amazingly effective and surprisingly robust against disease. But make no mistake, the _species_ among them also compete in evolutionary terms and die out or migrate as environments shift.
People are trained to think that extinctions destroy life because they _do_ destroy ecosystems in wholesale and clearly destructive fashions, in the short term. Mass extinctions cause change, certainly, and shake limbs of the evolutionary tree in ways that knock off branches and open gaps for new ones. But it's often a mistake to think of it as being automatically progress in other ways than merely evolutionary survival. Both knowledge and commercial possibilities are lost.
Take the example of the American bison: nearly extinct from human predation, the bison makes more meat, and leaner, more cardiac friendly meat, with less arable land and water than cows. They're a far better meat animal, and are only recently being repopulated from the almost too-small-to-safely-breed herds that were left. Or take the pigeon population in cities: pigeons are basically flying rats, carrying infection and dropping their feces on human heads below, and destroying roofs and gutters and wires quite effectively. Re-introducing hawks into cities has been very effective in controlling them.
Similar ecological and commercial niches are accidentally destroyed when mass extinction occurs, and it is certainly a loss to species (like us) who find those niches useful.
Oh. Oh, my. No, the "successful" ones are suited that particular biosphere. Change the biosphere, and other species will prosper in the short term, and entirely new blances form. Look at the introduction of plants to the biosphere and the introduction of this poisonous, highly chemically active gas "oxygen" to see how entirely new balances had to be formed.
Precise? Hardly. Color blindness, albinism, dwarfism, hay fever, and diabetes would have been left out of the hominid gene pools thousands of generations ago if this were true. No, evolution is very casual surgery indeed, and the scars of amputations and major grafts can be traced throughout the evolutionary tree.
Amusingly, Slashdot is choking on the lengthy examples I'm trying to quote, complaining about excessively long strings.
If you do all your coding by cut & paste inside a GUI, I suppose you don't care. But the "_" is far easier to read and differentiate elements in. And I've wasted far, far too much time in the last few years with careless programmers misusing such confusing syntax for mistyped local versions of functions, propagating them into different levels of the Java hierarchies, and having to go clean it up later. The tendency to mistype, to miss plurals, and to alternatively use or not use articles or to fail to capitalize them wastes a lot of programmer time.
Yes, they would. Fragmenting the words is far more legible than incredibly long mixed case, even thought the _ is a bit harder on the typist. And the fragmentation encourages consistency of grouping libraries of functions in a more visible fashion.
C coding standards don't encourage it as a standard. TheMixedCapitalizationReallyDoesHurtProgrammersHands, and ItEncouragessSpellingIssuessForPeopleWhoCantCountExcesLetterSes.
Now find the spelling errors in my function names.
IBM was right to lowball. Sun, as innovative as it was, had become famous for failed technology leadership efforts that drifted into fiscal blackholes. Their Sparc architecture has drifted out of engineering use, their compilers were proprietary and unstable and the first thing to replace on Sun systems, their dalliance with AT&T based UNIX rather than BSD turned into a horrific waste of time in competition with Linux's plethora of tools and ongoing development, and even their attempts at new BIOS technologies were hamstrung by their own incapacity to install new kernels on their own systems. I had to ship back a rack full of those Cobalt servers because _no one_ in Sun support could help me get new kernels on them, and they were unable to let me talk to a real engineer who'd ever actually done it rather than merely someone with a "How-To" guide they couldn't even ship me a text copy of.
The larger gambling businesses will _never_ be within reach of their righteous neighbors: both of them want buffer zones for their own protection. Look at Las Vegas and Atlantic City, and more recently at the spread of drugs, prostitution, and loan sharking with the Indian casinos. Casinos, like strip clubs, do not create safe neighborhoods and are usually isolated from businesses and residential neighborhoods. (They do tend to have lively theater and restaurant districts nearby.)
Given the amazingly poor security of many law enforcement and medical databases, wouldn't it be easier to simply swap the data with some innocent sucker?
Oh dear, I wish they would go away. They're as bad as those street vendors selling "genuine Rolex" watches for $20. You literally do not know what motherboards, CPU's, or controllers those vendors have used. Opening one up and finding that the drives behave erratically because the RAID controller is some unlabeled card whose chips have had the numbers scraped off is a serious problem. And calling one to get the known BIOS update for the motherboard is an amazing adventure in being asked "have you rebooted your Windows machine? please open the control panel and look up this information" that is a complete waste of time for Linux admins.
And notice precisely how much Microsoft funded them: enough to continue the lawsuits, but not enough to actually allow them to continue development with what had been a fairly robust version of UNIX and some interesting SMP features.
No, for ordinary customer use, read the license. (http://www.nvidia.com/content/DriverDownload-March2009/licence.php?lang=us). Tha't the normal, single user license. A commercial vendor might arrange a different license, perhaps associated with a bulk purchase: this vendor did not bother to do so.
This NVidia license also does not eliminate the GPL on the Linux kernel, which you can review at http://www.gnu.org/licenses/licenses.html. The GPL and the "tainting" of the GPL kernel by the NVidia modules prevents vendors shipping it pre-installed. That means no OS images can be shipped with it pre-installed, it has to be installed on a system by system basis.
At my last look, SuSE came with a badly written YaST utility that would reach out to Nvidia and download a frankly out of date NVidia. This isn't the same thing at all as being able to pre-install and preship the machines with the drivers ready to go.
"Replacing the default Mesa ones" is fine. Doing it in complete secrecy from the operating system's package management, so that the ordinary vendor provided updates will overwrite the NVidia drivers, is both foolish and and destabilizing.
I'm delighted for friend that he's doing well. Who set up the drivers for him, or did he do it himself? The licensing can be quite awkward, and to be frank, a lot of users don't pay attention to it.
I know at least 200 CGI artists whose IT department would love to switch to Linux and use economically affordable but quite powerful NVidia cards, and a desktop vendor who lost the sale because they couldn't legally pre-install the NVidia drivers nor rely on the NVidia setups to remain stable. The NVidia installer moves aside OpenGL libraries and replaces them: any software updates that accidentally include fresh OpenGL libraries break the NVidia setup.
They're testing ATI based video cards right now to try and close the deal.
> Having more pixels is a good thing for anyone who sells flash memory.
Here, I fixed that for you.
It's true that many photos would be improved by more detail. But it's not always a benefit: just as text is well-represented with a modest number of bits to describe a letter in ASCII, storing sophisticated graphical images of each character is usually quite pointless and actually interferes with getting work done.
Take a good look at https://www.firstcitizens.com/privacy_security/tips_avoid_cashiers_check_fraud.html: This is First Citizen's bank, a very popular US bank, explaining how this fraud works. As far as you as a normal client can tell, and as far as the bank will tell you, the check has "cleared", but it can still be blocked by the other bank for lack of funds.
Did you compile it yourself, on a no longer available operating system and a non-gcc based compiler? Can you do it _again_, now 30 or more years later?
It's useless for coal because coal is typically more than 10 half-lives of C14 old. The longest estimates for C14 accuracy seem to range up to 60,000 years (mentioned over on Wikipedia).
Slashdot Mirror
You're quite correct, there are uses for it. Protocols that have sufficient inherent redundancy and error recovery are reasonable for this kind of broadcast. But I've run into a number of instances where some "clever" engineer said "just use multicast and we won't need so many ports and switches!", especially for data where missed packets was considered a failure and unacceptable. Then the job was left for the next few years of failures and debugging. By the time they'd finished laying proprietary, unstable, and unsupportable checksumming and packet indexing and creative mishandling of their networks, they'd spent far more on lost services and engineering time than if they'd simply run TCP from the start.
In many instances, it's more robust to scale the TCP: a cautiously built architecture without excessive bandwidth on individual channels can scale quite well, with very modest devices acting as proxies for the TCP traffic. But, sadly, that doesn't seem to be a very "exciting" technology, and isn't prey to the optimistic underbudgeting I've seen with multicast projects, and the tendency to assume that what works with 20% of the expected on only 2 hosts in the same rack running entirely hand-built test software will work with dozens or hundreds of machines scattered to multiple data centers with divergent platforms and services on them.
Similar to the expansion of the US "wild west", we're due for years of backfilling and territory arguments. Look ahead to the owners of /8 address ranges having them confiscated. (MIT, for example, hardly needs it: they should be NAT'ing all their internal traffic anyway to prevent "computer science majors" from pulling stupid stunts like the David LaMacchia case (http://www.informit.com/articles/article.aspx?p=169520).
NAT is notoriously lighter weight to support than IPv6, and helps provide some border control of undesirable services from inside your network. Replacing the router infrastructure and the configuration tools for stable, legacy systems to support IPv6 is expensive and the benefits of IPv6 are frankly underwhelming. It's exciting "auto-configuration" is, in most cases, a horrendously bad idea for public facing systems, and private systems don't need it. Useful security features, such as IPsec, were backported to IPv4. And the robust technical features of IPsec seem to be overwhelmed by the far easier to use client behavior of PPTP.
Multicast? Oh, dear. Do _not_ get me started on the flaws of multicast programmers decided that the lack of information about missed packets in multicast forcing them to rewrite TCP, badly, as an unstable software layer on top of multicast.
Other folks will mention their favorite virtualization toolkits. I'd like to mention that if the students expect to do serious work, such as rebuilding interesting packages to get their feet wet, they may well need more than 4 Gig of space. They may need swap space if their laptop or host machine is underpowered, and after you get done installing compilers and development libraries and Gnome or KDE, there's surprisingly little space left on a 4 Gig disk. Also, if they attempt to build RPM's in the chrooted "mock" or "mach" environments for stability of build environments, that's another chunk of space.
So if your budget can stand it, try to get 8 Gig USB sticks.
New York City and LA are interesting issues. One problem is the conflict between state and local authorities: the state expects state wide coverage, and high bandwidth providers have zero interest in building and maintaining high speed infrastructure to rural areas: it's quite expensive.
Another is that there is a piranha-like feeding frenzy going on for network bandwidth in such core urban areas: the companies are interfering with each other, both inside and outside the courts and lobbyist's offices, to _block_ progress by their competitors. And the arguments range from profound (what role should government have in using eminent domain to allow fiber runs or cell phone towers) to completely ridiculous. (How dare a provider tell me I can't run my own SMTP server! I'll send any email I want, how dare you call it spam, you're infringing on my freedom of speech!)
Fair enough. Sweden, rather than being _small_, is merely modest. Yes, it's a large country for Europe, but it's still a pretty modest size. I'd be curious how they fund their network infrastructure.
The Netherlands are _small_, less than 15,000 squarae miles, densely populated and highly developed. You have to be, in order to keep pushing back the ocean every year. (And good luck with that if global warming makes the oceans rise noticeably!) It is much easier to build dense, effective infrastructure to support such bandwidth in such a small area than it is in the USA, which has not only federal issues but 50 different state governments to negotiate with, and the states have vastly different requirements.
But cockroaches are not one species, any more than primates are all one species. They are an "order" a set of species with some common characteristics. Like sharks, they're amazingly effective and surprisingly robust against disease. But make no mistake, the _species_ among them also compete in evolutionary terms and die out or migrate as environments shift.
People are trained to think that extinctions destroy life because they _do_ destroy ecosystems in wholesale and clearly destructive fashions, in the short term. Mass extinctions cause change, certainly, and shake limbs of the evolutionary tree in ways that knock off branches and open gaps for new ones. But it's often a mistake to think of it as being automatically progress in other ways than merely evolutionary survival. Both knowledge and commercial possibilities are lost.
Take the example of the American bison: nearly extinct from human predation, the bison makes more meat, and leaner, more cardiac friendly meat, with less arable land and water than cows. They're a far better meat animal, and are only recently being repopulated from the almost too-small-to-safely-breed herds that were left. Or take the pigeon population in cities: pigeons are basically flying rats, carrying infection and dropping their feces on human heads below, and destroying roofs and gutters and wires quite effectively. Re-introducing hawks into cities has been very effective in controlling them.
Similar ecological and commercial niches are accidentally destroyed when mass extinction occurs, and it is certainly a loss to species (like us) who find those niches useful.
Oh. Oh, my. No, the "successful" ones are suited that particular biosphere. Change the biosphere, and other species will prosper in the short term, and entirely new blances form. Look at the introduction of plants to the biosphere and the introduction of this poisonous, highly chemically active gas "oxygen" to see how entirely new balances had to be formed.
Precise? Hardly. Color blindness, albinism, dwarfism, hay fever, and diabetes would have been left out of the hominid gene pools thousands of generations ago if this were true. No, evolution is very casual surgery indeed, and the scars of amputations and major grafts can be traced throughout the evolutionary tree.
Amusingly, Slashdot is choking on the lengthy examples I'm trying to quote, complaining about excessively long strings.
If you do all your coding by cut & paste inside a GUI, I suppose you don't care. But the "_" is far easier to read and differentiate elements in. And I've wasted far, far too much time in the last few years with careless programmers misusing such confusing syntax for mistyped local versions of functions, propagating them into different levels of the Java hierarchies, and having to go clean it up later. The tendency to mistype, to miss plurals, and to alternatively use or not use articles or to fail to capitalize them wastes a lot of programmer time.
Yes, they would. Fragmenting the words is far more legible than incredibly long mixed case, even thought the _ is a bit harder on the typist. And the fragmentation encourages consistency of grouping libraries of functions in a more visible fashion.
C coding standards don't encourage it as a standard. TheMixedCapitalizationReallyDoesHurtProgrammersHands, and ItEncouragessSpellingIssuessForPeopleWhoCantCountExcesLetterSes.
Now find the spelling errors in my function names.
IBM was right to lowball. Sun, as innovative as it was, had become famous for failed technology leadership efforts that drifted into fiscal blackholes. Their Sparc architecture has drifted out of engineering use, their compilers were proprietary and unstable and the first thing to replace on Sun systems, their dalliance with AT&T based UNIX rather than BSD turned into a horrific waste of time in competition with Linux's plethora of tools and ongoing development, and even their attempts at new BIOS technologies were hamstrung by their own incapacity to install new kernels on their own systems. I had to ship back a rack full of those Cobalt servers because _no one_ in Sun support could help me get new kernels on them, and they were unable to let me talk to a real engineer who'd ever actually done it rather than merely someone with a "How-To" guide they couldn't even ship me a text copy of.
The larger gambling businesses will _never_ be within reach of their righteous neighbors: both of them want buffer zones for their own protection. Look at Las Vegas and Atlantic City, and more recently at the spread of drugs, prostitution, and loan sharking with the Indian casinos. Casinos, like strip clubs, do not create safe neighborhoods and are usually isolated from businesses and residential neighborhoods. (They do tend to have lively theater and restaurant districts nearby.)
Given the amazingly poor security of many law enforcement and medical databases, wouldn't it be easier to simply swap the data with some innocent sucker?
Oh dear, I wish they would go away. They're as bad as those street vendors selling "genuine Rolex" watches for $20. You literally do not know what motherboards, CPU's, or controllers those vendors have used. Opening one up and finding that the drives behave erratically because the RAID controller is some unlabeled card whose chips have had the numbers scraped off is a serious problem. And calling one to get the known BIOS update for the motherboard is an amazing adventure in being asked "have you rebooted your Windows machine? please open the control panel and look up this information" that is a complete waste of time for Linux admins.
And notice precisely how much Microsoft funded them: enough to continue the lawsuits, but not enough to actually allow them to continue development with what had been a fairly robust version of UNIX and some interesting SMP features.
No, for ordinary customer use, read the license. (http://www.nvidia.com/content/DriverDownload-March2009/licence.php?lang=us). Tha't the normal, single user license. A commercial vendor might arrange a different license, perhaps associated with a bulk purchase: this vendor did not bother to do so.
This NVidia license also does not eliminate the GPL on the Linux kernel, which you can review at http://www.gnu.org/licenses/licenses.html. The GPL and the "tainting" of the GPL kernel by the NVidia modules prevents vendors shipping it pre-installed. That means no OS images can be shipped with it pre-installed, it has to be installed on a system by system basis.
At my last look, SuSE came with a badly written YaST utility that would reach out to Nvidia and download a frankly out of date NVidia. This isn't the same thing at all as being able to pre-install and preship the machines with the drivers ready to go.
"Replacing the default Mesa ones" is fine. Doing it in complete secrecy from the operating system's package management, so that the ordinary vendor provided updates will overwrite the NVidia drivers, is both foolish and and destabilizing.
I'm delighted for friend that he's doing well. Who set up the drivers for him, or did he do it himself? The licensing can be quite awkward, and to be frank, a lot of users don't pay attention to it.
I know at least 200 CGI artists whose IT department would love to switch to Linux and use economically affordable but quite powerful NVidia cards, and a desktop vendor who lost the sale because they couldn't legally pre-install the NVidia drivers nor rely on the NVidia setups to remain stable. The NVidia installer moves aside OpenGL libraries and replaces them: any software updates that accidentally include fresh OpenGL libraries break the NVidia setup.
They're testing ATI based video cards right now to try and close the deal.
Middle management.
I'm sure there are plenty of expatriate Tibetans who'd be happy to take those jobs.
> Having more pixels is a good thing for anyone who sells flash memory.
Here, I fixed that for you.
It's true that many photos would be improved by more detail. But it's not always a benefit: just as text is well-represented with a modest number of bits to describe a letter in ASCII, storing sophisticated graphical images of each character is usually quite pointless and actually interferes with getting work done.
Take a good look at https://www.firstcitizens.com/privacy_security/tips_avoid_cashiers_check_fraud.html: This is First Citizen's bank, a very popular US bank, explaining how this fraud works. As far as you as a normal client can tell, and as far as the bank will tell you, the check has "cleared", but it can still be blocked by the other bank for lack of funds.
Did you compile it yourself, on a no longer available operating system and a non-gcc based compiler? Can you do it _again_, now 30 or more years later?
Good luck with that. There some ancient binaries and source code at ftp://kermit.columbia.edu/kermit/bin/ and ftp://kermit.columbia.edu/kermit/old/. But it wouldn't be clear to me that our original poster has the tools to install it if it's not already present.
It's useless for coal because coal is typically more than 10 half-lives of C14 old. The longest estimates for C14 accuracy seem to range up to 60,000 years (mentioned over on Wikipedia).