Basically, everyone who wants one can have a static address.
...
There aren't really any downs to IPv6 other than the replacement costs. Possibly privacy issues -- there's been interest in using your MAC address as the last bits of your IPv6 address, which seems incredibly stupid to me -- like one huge, protocol-independent, world-readable cookie, but whatever.
In which way is a static address not a huge, protocol-independent, world-readable cookie? In Denmark, cable modem users get mostly static addresses; the DHCP server will keep handing out the same address to the same MAC (The address will only change when the provider renumbers. It happens, but rarely.). I must admit it does not disturb my sleep that I have this number fixed to my brow, err, my Internet traffic. Of course there are ways to get a new address from DHCP, most obviously changing the MAC address. Guess what that will do to an IPv6 address derived from the MAC...
Besides, you can just go through a proxy, and the "cookie" will disappear. (Oh by the way, I run 6to4 on the gateway and stateless autoconfiguration on the other machines, so my IPv6 traffic can be identified both by my static IPv4 address and the MAC addresses of my various machines. The horror.)
One of the big problems with IPv4 is that worms can trivially scan the complete address space. With IPv6 that is not practical. This means that worms would have to use other methods, such as guessing dns names and resolving them to IPv6 addresses. This would slow them down tremendously and cause them to fail to hit most of the vulnerable machines. In contrast, Code Red managed to get behind firewalls in many companies. To me it looks like the IPv6 scenario is safer to a naive user (the kind who thinks that NAT protects them), and any security policy that is applied to IPv4 can be applied equally well to IPv6.
If you happen to have at one IPv4 address, you are automatically allocated a/48 subnet on IPv6 with 6to4. For free. Good luck trying to run out of addresses (for the non-initiated, a/48 contains 2^80 addresses).
This article is unnecessarily alarming, but then again, who would bother reading an article with this headline: "6bone users have to change addresses in three years"?.
Chemical rockets are generally fine. The big problem is that they carry both fuel and oxidizer. Oxygen is heavy compared to the fuel, and during the first part of the trip (where most of the fuel is burned) there is plenty of oxygen available for free right outside the rocket.
An air-breathing first stage would be much lighter. Designing an airbreathing engine that works from subsonic to mach-10 or mach-20 has sadly proven somewhat challenging.
Ok, imagine your company makes sensors that output their results via RS-232 serial. Or controllers that are given commands via RS-232 serial. Or maybe you have machinery that is programmed through RS-232 serial. Either way, your would like to access those products remotely, and RS-232 just doesn't go very far. Add this thing, and suddenly your products are web-enabled.
The price is a bit high still, but there is a lot of equipment where $33 extra a unit would not scare customers away.
You are assuming that the market is elastic. As far as I can tell, the market for copyrighted works is very inelastic. If the market was elastic, the producers would be encouraged to lower prices, since they earn the most when the prices are near the variable cost. In an inelastic market, rising prices lead to higher profits.
The obvious choice is to get a second set of 16 addresses on the other connection, and then make your DNS server send out addresses to whichever connection currently works. Not all services like switching addresses, and sessions break when doing failover, but it might work for you. If you only care about outgoing traffic, load-balancing and failover is fairly easy to do and there are lots of products to help. Again, outgoing sessions will get killed if they happen to use the link that breaks.
To ensure that we do not hit the address space problems again, most of the IPv6 space has been set aside as reserved. I will quote RFC 3117:
We are highly confident in the validity of this analysis, based on experience with IPv4 and several other address spaces, and on extremely ambitious scaling goals for the Internet amounting to an 80 bit address space *per person*. Even so, being acutely aware of the history of under-estimating demand, the IETF has reserved more than 85% of the address space (i.e., the bulk of the space not under the 001 Global Unicast Address prefix). Therefore, if the analysis does one day turn out to be wrong, our successors will still have the option of imposing much more restrictive allocation policies on the remaining 85%.
MSSQL has holes like a swiss cheese. Furthermore, applying hotfixes to the thing is a nightmare involving, among other things, shutting it down and moving files around by hand. Not something you really want to do on a hundred servers every few weeks. As of SQL Server 2000 Service Pack 2 it can be crashed with a single-byte UDP packet. Never expose any of the SQL Server ports on the Internet. Blocking TCP port 1434 is a good start, but not sufficient.
Do note that just blindly switching to Oracle will not necessarily help. All products that are usually connected to the Internet, such as sendmail or IIS, have had the crap beat out of them by now and are relatively secure (though both sendmail and IIS are secure despite their architecture, not because of it). Now the time has come for those products that are usually not placed in such a hostile environment. Oracle realized this a bit before the MSSQL people did, but both products still have a way to go.
I invite you to get out your meter and verify the results independently before accusing me of spouting BS!:) I don't claim to be a power supply design expert, but I have fried enough of them to learn a little bit about their characteristics.
It is practically impossible for non-professionals to measure the actual power used by computers these days. The common devices you can buy (or around here, borrow from the electric company) will only measure correctly when the load is purely resistant. Switching power supplies confuse them greatly.
Whether the meter used for billing electricity measures correctly is doubtful too. I would bet against it. However, it is only fair that people pay more, since the electricity company has to compensate. Anyway, legislation is coming into effect in the EU requiring that all new devices have to present a load that is almost purely resistant. This will make power supplies more expensive, but reduce losses.
Incidentally, all this would be a lot easier if we just forgot that old-fashioned AC stuff and switched to proper DC.
Support may cease anytime, and continued development is uncertain. With Free Software you can pay someone to fix problems when the original developer is gone. With shareware, you are screwed.
It is not that unusual to see two netcards with the same MAC -- most often 00:00:00:00:00:00. Usually the problem is either that the EEPROM was not properly initialized from the factory, or that some driver figured it should override what the EEPROM says. It is a really annoying problem to diagnose if you don't get the right idea quickly, because most things still work most of the time. IP is very resilient, and some misdirected packets won't completely break anything.
Really big buffers, or maybe temp files behind the scenes? There's technically nothing about the concept of a pipe that requires multitasking. It's just really painful without it.
Of course that depends on what you mean by "pipe". This one would be slightly difficult when single tasking (unless you happen to have an infinite buffer lying around somewhere):
yes | rm/dev/*
These days yes is not needed very much, since most commands have a force switch, but in the past it was very useful.
1: All power usage on Earth is converted to electricity.
2: The historic growth trend continues.
3: All of this growing electrical generation need comes from tidal energy.
You cannot usefully project the current growth trend very far. First, energy consumption in 2000 averaged 1.3e13W.
Assume that energy consumption doubles every 50 years, which is a growth of around 1.4% a year. Then in 1000 years energy use is about a million (2^20) times as large, or 1.3e19W. The energy received by Earth from the Sun is around 2e17W. The temperature of the Earth would have to increase a lot in order to radiate all that waste heat away.
Now jump to just 10000 years, when energy consumption has increased 1.6E60 times to 2e79W. The Sun outputs somewhere around 4e26W.
5e46 new stars will have to be acquired.
Talking about millions of years just gets absurd. The historic trend will have to change.
In reality the beam would be no more powerful then a cell phone.
The idea is to get more energy per area down on Earth than direct solar power can get you. Cell phone radiation even right next to the antenna is way less powerful than solar radiation.
Not that I'm usually a tree-hugger, but it strikes me as hypocritical that the Scandinavians come across as looking good for pursuing "alternative" energy, when in fact that pursuit is motivated by profit margin and a scarcity of fossil fuel.
Yes, Norway really wants lower oil prices so they can spend less on all that fossil fuel imported there. Oh wait, Norway exported 34.62 million tons of crude oil just in the second quarter of 2002.
People who contribute code to Linux (the kernel) do not give up copyright on that code. It can be debated whether contributing without making a clear statement about the intended license implies that the GPL-with-module-exception or just plain GPL applies. However, several contributors have clearly stated that their work is only available under the unmodified GPL. There is no ambiguity about that. Their code was accepted into the kernel knowing what their licensing choice was.
The licensing is between the contributor and the licensee. Linus Torvalds is not a part in this contract (except for the parts he wrote). Of course someone could try to say they were led to believe that a different license applied, since Linus Torvalds included the blurb about the module exception, but either way blaming Linus will most likely not get them out of an injunction against distribution of their software. Whether the gains from binary modules offset the risk of legal troubles will have to be decided by each company, of course.
Another thing to note is that not all parts of Linux (the kernel) were contributed directly to it. Some parts were developed for other projects and then they were included in Linux because their licenses allow it. It seems obvious that the module exception cannot be applied to such entirely unrelated code.
The fact that NVIDIA and Rational have not been sued does not imply that they are legally in the clear. There are certainly several kernel developers who have said clearly that the "module exception" invented by Linus Torvalds does not extend to linking to their code. So far they have not sued anyone. Maybe they never will.
Re:Multiple universes?
on
One of Many
·
· Score: 2
Yes and no - MW lists three meanings for atom, only the third of which describes the atom as a particle (which may be split). It's true that the other senses of the word preclude splitting any further, but they are other senses of the word.
The best online explanation of the etymology of the word atom seems to be at a Geocities page. I cannot believe I just made a link there. (If you see me recommending AOL and hotmail, please have me committed.)
Re:Multiple universes?
on
One of Many
·
· Score: 5, Insightful
This is one of my pet hates. By the very definition of the word [m-w.com], there can only be one universe. Or are the definitions now being changed?
By the very nature of the word, an atom cannot be divided.
Slashdot Mirror
There aren't really any downs to IPv6 other than the replacement costs. Possibly privacy issues -- there's been interest in using your MAC address as the last bits of your IPv6 address, which seems incredibly stupid to me -- like one huge, protocol-independent, world-readable cookie, but whatever.
In which way is a static address not a huge, protocol-independent, world-readable cookie? In Denmark, cable modem users get mostly static addresses; the DHCP server will keep handing out the same address to the same MAC (The address will only change when the provider renumbers. It happens, but rarely.). I must admit it does not disturb my sleep that I have this number fixed to my brow, err, my Internet traffic. Of course there are ways to get a new address from DHCP, most obviously changing the MAC address. Guess what that will do to an IPv6 address derived from the MAC...
Besides, you can just go through a proxy, and the "cookie" will disappear. (Oh by the way, I run 6to4 on the gateway and stateless autoconfiguration on the other machines, so my IPv6 traffic can be identified both by my static IPv4 address and the MAC addresses of my various machines. The horror.)
One of the big problems with IPv4 is that worms can trivially scan the complete address space. With IPv6 that is not practical. This means that worms would have to use other methods, such as guessing dns names and resolving them to IPv6 addresses. This would slow them down tremendously and cause them to fail to hit most of the vulnerable machines. In contrast, Code Red managed to get behind firewalls in many companies. To me it looks like the IPv6 scenario is safer to a naive user (the kind who thinks that NAT protects them), and any security policy that is applied to IPv4 can be applied equally well to IPv6.
If you happen to have at one IPv4 address, you are automatically allocated a /48 subnet on IPv6 with 6to4. For free. Good luck trying to run out of addresses (for the non-initiated, a /48 contains 2^80 addresses).
This article is unnecessarily alarming, but then again, who would bother reading an article with this headline: "6bone users have to change addresses in three years"?.
Chemical rockets are generally fine. The big problem is that they carry both fuel and oxidizer. Oxygen is heavy compared to the fuel, and during the first part of the trip (where most of the fuel is burned) there is plenty of oxygen available for free right outside the rocket.
An air-breathing first stage would be much lighter. Designing an airbreathing engine that works from subsonic to mach-10 or mach-20 has sadly proven somewhat challenging.
What about the hydrogen economy? Maybe someone should tell W that Jupiter is pretty much all hydrogen...
Ok, imagine your company makes sensors that output their results via RS-232 serial. Or controllers that are given commands via RS-232 serial. Or maybe you have machinery that is programmed through RS-232 serial. Either way, your would like to access those products remotely, and RS-232 just doesn't go very far. Add this thing, and suddenly your products are web-enabled.
The price is a bit high still, but there is a lot of equipment where $33 extra a unit would not scare customers away.
What makes you think another copy will not be produced for $2 and bought for $40 by that guy in the example?
You are assuming that the market is elastic. As far as I can tell, the market for copyrighted works is very inelastic. If the market was elastic, the producers would be encouraged to lower prices, since they earn the most when the prices are near the variable cost. In an inelastic market, rising prices lead to higher profits.
I think IBM did steal the "high degree of design coordination" from SCO. It is at least very clear that SCO doesn't have any such thing.
The "active ingredients" in a pencil are graphite and clay. Lead is no longer used.
The obvious choice is to get a second set of 16 addresses on the other connection, and then make your DNS server send out addresses to whichever connection currently works. Not all services like switching addresses, and sessions break when doing failover, but it might work for you. If you only care about outgoing traffic, load-balancing and failover is fairly easy to do and there are lots of products to help. Again, outgoing sessions will get killed if they happen to use the link that breaks.
MSSQL has holes like a swiss cheese. Furthermore, applying hotfixes to the thing is a nightmare involving, among other things, shutting it down and moving files around by hand. Not something you really want to do on a hundred servers every few weeks. As of SQL Server 2000 Service Pack 2 it can be crashed with a single-byte UDP packet. Never expose any of the SQL Server ports on the Internet. Blocking TCP port 1434 is a good start, but not sufficient.
Do note that just blindly switching to Oracle will not necessarily help. All products that are usually connected to the Internet, such as sendmail or IIS, have had the crap beat out of them by now and are relatively secure (though both sendmail and IIS are secure despite their architecture, not because of it). Now the time has come for those products that are usually not placed in such a hostile environment. Oracle realized this a bit before the MSSQL people did, but both products still have a way to go.
It is practically impossible for non-professionals to measure the actual power used by computers these days. The common devices you can buy (or around here, borrow from the electric company) will only measure correctly when the load is purely resistant. Switching power supplies confuse them greatly.
Whether the meter used for billing electricity measures correctly is doubtful too. I would bet against it. However, it is only fair that people pay more, since the electricity company has to compensate. Anyway, legislation is coming into effect in the EU requiring that all new devices have to present a load that is almost purely resistant. This will make power supplies more expensive, but reduce losses.
Incidentally, all this would be a lot easier if we just forgot that old-fashioned AC stuff and switched to proper DC.
Support may cease anytime, and continued development is uncertain. With Free Software you can pay someone to fix problems when the original developer is gone. With shareware, you are screwed.
It is not that unusual to see two netcards with the same MAC -- most often 00:00:00:00:00:00. Usually the problem is either that the EEPROM was not properly initialized from the factory, or that some driver figured it should override what the EEPROM says. It is a really annoying problem to diagnose if you don't get the right idea quickly, because most things still work most of the time. IP is very resilient, and some misdirected packets won't completely break anything.
Of course that depends on what you mean by "pipe". This one would be slightly difficult when single tasking (unless you happen to have an infinite buffer lying around somewhere):
yes | rm /dev/*
These days yes is not needed very much, since most commands have a force switch, but in the past it was very useful.
2: The historic growth trend continues.
3: All of this growing electrical generation need comes from tidal energy.
You cannot usefully project the current growth trend very far. First, energy consumption in 2000 averaged 1.3e13W. Assume that energy consumption doubles every 50 years, which is a growth of around 1.4% a year. Then in 1000 years energy use is about a million (2^20) times as large, or 1.3e19W. The energy received by Earth from the Sun is around 2e17W. The temperature of the Earth would have to increase a lot in order to radiate all that waste heat away.
Now jump to just 10000 years, when energy consumption has increased 1.6E60 times to 2e79W. The Sun outputs somewhere around 4e26W. 5e46 new stars will have to be acquired.
Talking about millions of years just gets absurd. The historic trend will have to change.
The idea is to get more energy per area down on Earth than direct solar power can get you. Cell phone radiation even right next to the antenna is way less powerful than solar radiation.
Yes, Norway really wants lower oil prices so they can spend less on all that fossil fuel imported there. Oh wait, Norway exported 34.62 million tons of crude oil just in the second quarter of 2002.
The licensing is between the contributor and the licensee. Linus Torvalds is not a part in this contract (except for the parts he wrote). Of course someone could try to say they were led to believe that a different license applied, since Linus Torvalds included the blurb about the module exception, but either way blaming Linus will most likely not get them out of an injunction against distribution of their software. Whether the gains from binary modules offset the risk of legal troubles will have to be decided by each company, of course.
Another thing to note is that not all parts of Linux (the kernel) were contributed directly to it. Some parts were developed for other projects and then they were included in Linux because their licenses allow it. It seems obvious that the module exception cannot be applied to such entirely unrelated code.
The fact that NVIDIA and Rational have not been sued does not imply that they are legally in the clear. There are certainly several kernel developers who have said clearly that the "module exception" invented by Linus Torvalds does not extend to linking to their code. So far they have not sued anyone. Maybe they never will.
The best online explanation of the etymology of the word atom seems to be at a Geocities page. I cannot believe I just made a link there. (If you see me recommending AOL and hotmail, please have me committed.)
By the very nature of the word, an atom cannot be divided.