There is another assumption that may not hold (but usually does).
The code assumes that a writes are atomic. This will almost always hold for 8 processes and usually form 32, but if the flags array is larger than a word, atomic writes go out the window.
Second that. I've got an Antec riser with a built-in fan. Runs off of a usb port.
It has done wonders for my PowerBook. I run it with the lid closed and an external monitor sometimes and it used to get really hot when I closed the lid, but the extra fan does a great job.
NAT is definitely NOT BETTER than IPv6. That is not to say that IPv6 is better. They are two different technologies that operate in two different ways and solve different (but intersecting) sets of problems.
However, it does mitigate the address space problems. Basically, the reason why it prevents attacks is that when home users put a NAT router between the internet and their machine, the nature of NAT means that they also have the equivalent of a sanely configured firewall there as well. Same goes for larger scale NAT, which is less common.
I may be wrong, but not having the real IP address is not where the real advantage comes in, it comes from the fact that internet hosts cannot connect to you without you contacting them first. Any well configured firewall (network or host) will do the same thing.
IPv6 provides many other benefits, and some drawbacks.
I thought everyone knew about this, but the same are available in linux as/dev/random and/dev/urandom.
random is completely random and will only give out numbers until it runs out of entropy. urandom continues to hand out numbers even if the entropy of the kernels pool has reached zero.
Of course any use of these interfaces assumes trust that your OS hasn't been compromised (not a big deal the same is required when you use the computer to encrypt something) and that the implementors did things properly. I can't speak for the BSD implementation, but the linux implementation has been given quite a bit of thought and is well documented (of course I'm not a cryptographic expert, though).
It could be admitted as evidence that he had the means to commit the crime, but not as evidence of intent, which is the issue at hand. Having a video studio in your house does not show intent to shoot kiddy porn. It shows intent to shoot a movie.
If there is no other use for a tool or it was actually used to cover up a crime, then it shows intent. I own a mop. If I shot an intruder in my home, does that mop show that I intended to murder the victim because I have something to cover it up.
If the police have encrypted files that contain videos, that shows intent, not having encryption software.
Actually the 5th amendment probably wouldn't protect you in this case. If the police have a search warrant that includes the contents of those encrypted files you would have to turn over the key the same way you would have to turn over the key to your safe if they have a warrant.
The good password requiremnt is not helped by the fact that users are also required to change it every xx days, so not only you need to remember a strange password, you have to remember a different one every couple of days
Yeah, plus having to buy all that new hardware gets expensive!
>Any other requirements could be done by stepping voltage down.h
You can't just step down DC power, that is one of the many reasons AC is ubiquitous when most electronics operate on DC.
With AC, voltages can be stepped down or up.
AC can be more easily transmitted over longer distances because it is transmitted at very high voltage(and therefore low current), then transformed near the destination to 110/220/whatever you might use.
AC can easily be converted into DC. Give me a good diode. Or even better, 4 diodes and a capacitor. The inverse it not true.
"booting" from a mirrored volume works just fine with most distros. If you are running custom kernel, you should be careful to make sure to compile in all the needed drivers, but if you're running a raid system you would probably do that anyway.
The only real restrictions are that/boot must be on a non-raid or mirrored raid partition. This is because the bios and boot loader need access to that heirarchy. Other than that you can software raid anything. The reason you can use a mirrored partition is that bios and grub treat it as read only and since it's mirrored it looks just like a normal partition to them.
I have been running small non-raid/boot with a striped root partition under Fedora for some time and haven't seen any problems.
I raid5 5 external scsi drives. When I want to swap on out, unplug it, plug in the new one, add it to the raid set and your done. No downtime. Simple. Checking/proc/mdstat will tell you how long it will take to fully integrate the new drive.
A better option might be to have this taken care of in hardware like bad sectors on some hard drives. Just have the chip do self tests on itself in the spare cycles when it isn't being used.
This would probably be a good option for small, low performance chips or RAM, but probably not for anything like a CPU.
Adding redundancy for high performance chips would require either duplicate cores, one of which would be turned off, or increasing the size of a single core. Increasing the size of the core, however would lead to lower clock speeds and lower performance to let impulses propagate over the extra space.
I would guess, though that turning one of two cores off if it fails a test and selling the cpu as a single core chip will be standard practice when dual core chips go into mass production, much the same way that chips that fail at higher clock speeds are sold as slower chips.
Before you say it, I know some companies mass produce dual core chips now, but I'm thinking mass as in x86 scale, not Power scale.
I assume SMTP settings are separate because SMTP servers are separate from pop/imap/etc accounts. Even though they happen to usually be on the same machine, they provide separate services.
You set the SMTP server to be the one where you want to send email through.
There is a standalone svn server, but if your users have ssh to the server, ssh is probably the easiest option to deal with.
From the administrative point of view, there is another issue that I haven't seen mentioned that svn helps with. The whole thing is kept in a single file, so you don't have many of the problems with group ouwnership that cvs has. Creating a new file in the repository is just a db write, so there is no new file, where cvs would create a new rcs file, which would probably need its group changed before anyone can use it.
1) Give me an example of these problems you're referring to. I really can't think of any way that app bundles are inferior to the Unix way.
The most glaringly obvious example is that the current structure is made so that as many files as possible can be 1)mounted on read-only drives 2)shared across machines.
If you read the fhs, it makes these points very clear./usr can be mounted read-only and shared by many machines over a network, so updates only have to be done once. To do updates, the drive is remounted read-write temporarily and then reverted to read-only. Being read-only brings both performance and security gains. Sharing saves disk space on clients and centralizes administration./etc can (usually) be mounted read-only, except when doing config changes, but needs to be specific to a machine./var needs to be read/write and specific to a machine./home needs to be read/write, but may be shared.
2) What's wrong with giving the *users* of the system the easiest way of doing things and letting the Administrators or Developers, the people who KNOW computers, doing the troubleshooting? The users can't troubleshoot; Administrators and Developers can.
Right now, users are more than welcome to keep packages in their own directory if they want. They can use any system they want and either set their path or link from ~/bin.
The question I would have for someone proposing the each package is a self contained heirarchy solution is "What are the advantages of using these packages as opposed to using a good package management system?" With a good package manager, if you want to know what files are part of a pachage, ask the package manager. If you want to remove the package, ask the package manager to. If you want to know what is installed, ask the package manager.
I was thinking the same thing. Think about the cost of console games. You could buy an xbox or a ps2 and have $35 left over, or for another $10 you could get 2 game cubes.
That doesn't really incorporate the whole problem either.
Big Iron also wins for applications that parallelize well, but require lots of communication amoung the nodes.
The reality however is that the best tool for the jobs is highly dependant on the individual application and it's hard to give a vague classification of what will work best in what situation.
The other issues are not just value and what works best, but what gives the best value. A cluster may provide the same performance for less money, but require an extra 100 man-days to code the application, while a big iron server may cost more , but save you the 100 man-days of work. Additionally, there will usually be less work to be done replacing pieces of the system simpley because there are a smaller number of pieces to break and they are usually higher quality.
We (University of Kentucky) just moved away from having the ssn on student id's, but it is still the same as our student identification number (for registration, billing, etc).
The electoral college is in place and had lasted this long for a reason. It forces candidates to go after voters across the nation/gives all states a say in the election.
The college has the same effect as the congress. Smaller states are given the same number of senators so that they are represented nationally. Larger states are given more representatives because they have more people and pay more taxes. States are given electoral votes so that smaller states have a say.
Without the electoral college, candidates would simply pander to NY, LA, Chicago, and maybe a few other markets. They can then safely ignore the rest of the nation because a small victory in the large markets overcomes even the largest losses in the rest of the nation.
The other reason for using the electoral college is that it creates a consensus. This year was a big enough victory that it doesn't really matter, but four years ago, the popular vote was too close to call(Let it go.), but the electoral college presented a clear winner(eventually). I think that after some thought, people on both sides of the issue would agree that if the election is really that close, we are better off with either candidate as president than with the presidency vacant.
I agree that clarity comes in handly, but I think the main issues you may have are with the distributors.
I use Fedora (bring on the flames) at home and a Mac at work. Everything I can think of on the Mac is easy to find on my Fedora box. Whatever you have chosen for email is labeled email. web is "Firefox Web Browser". Text editor is "Text Editor".
As for programs like vi, gawk, sed, etc. anyone whoe needs to use or knows how to use any of these as well as grep, lex, bison, emacs knows the commands right off.
Actually, TCP increases exponentially until the first packet is dropped. It backs off to half, then increases linearly, until another packet is dropped, backs off to half...
This means that a TCP connection uses ~75% of the bandwidth available to it (after all this stabilizes). So if there is only a single tcp connection over a given length, it will be 75% full at best. However, the whole reason for doing a lot of this is to allow many connections to coexist. If you transmit as fast as possible, you will get the highest throughput possible, but you will end up with a lot of dropped packets and won't play nice with others.
Slashdot Mirror
There is another assumption that may not hold (but usually does).
The code assumes that a writes are atomic. This will almost always hold for 8 processes and usually form 32, but if the flags array is larger than a word, atomic writes go out the window.
Second that. I've got an Antec riser with a built-in fan. Runs off of a usb port.
It has done wonders for my PowerBook. I run it with the lid closed and an external monitor sometimes and it used to get really hot when I closed the lid, but the extra fan does a great job.
NAT is definitely NOT BETTER than IPv6. That is not to say that IPv6 is better. They are two different technologies that operate in two different ways and solve different (but intersecting) sets of problems.
However, it does mitigate the address space problems. Basically, the reason why it prevents attacks is that when home users put a NAT router between the internet and their machine, the nature of NAT means that they also have the equivalent of a sanely configured firewall there as well. Same goes for larger scale NAT, which is less common.
I may be wrong, but not having the real IP address is not where the real advantage comes in, it comes from the fact that internet hosts cannot connect to you without you contacting them first. Any well configured firewall (network or host) will do the same thing.
IPv6 provides many other benefits, and some drawbacks.
How exactly is a peanut butter sandwich combining two condiments?
...
Peanut butter - arguably a condiment
Bread - Definitely not a condiment
Now, if you meant to reference that most divine of sandwiches, the pb & j
I thought everyone knew about this, but the same are available in linux as /dev/random and /dev/urandom.
random is completely random and will only give out numbers until it runs out of entropy. urandom continues to hand out numbers even if the entropy of the kernels pool has reached zero.
Of course any use of these interfaces assumes trust that your OS hasn't been compromised (not a big deal the same is required when you use the computer to encrypt something) and that the implementors did things properly. I can't speak for the BSD implementation, but the linux implementation has been given quite a bit of thought and is well documented (of course I'm not a cryptographic expert, though).
It could be admitted as evidence that he had the means to commit the crime, but not as evidence of intent, which is the issue at hand. Having a video studio in your house does not show intent to shoot kiddy porn. It shows intent to shoot a movie.
If there is no other use for a tool or it was actually used to cover up a crime, then it shows intent. I own a mop. If I shot an intruder in my home, does that mop show that I intended to murder the victim because I have something to cover it up.
If the police have encrypted files that contain videos, that shows intent, not having encryption software.
Actually the 5th amendment probably wouldn't protect you in this case. If the police have a search warrant that includes the contents of those encrypted files you would have to turn over the key the same way you would have to turn over the key to your safe if they have a warrant.
The good password requiremnt is not helped by the fact that users are also required to change it every xx days, so not only you need to remember a strange password, you have to remember a different one every couple of days
Yeah, plus having to buy all that new hardware gets expensive!
An emacs post replying to an Eclipse post and you choose to point out how bloated emacs is?
You can't just step down DC power, that is one of the many reasons AC is ubiquitous when most electronics operate on DC.
"booting" from a mirrored volume works just fine with most distros. If you are running custom kernel, you should be careful to make sure to compile in all the needed drivers, but if you're running a raid system you would probably do that anyway.
/boot must be on a non-raid or mirrored raid partition. This is because the bios and boot loader need access to that heirarchy. Other than that you can software raid anything. The reason you can use a mirrored partition is that bios and grub treat it as read only and since it's mirrored it looks just like a normal partition to them.
/boot with a striped root partition under Fedora for some time and haven't seen any problems.
The only real restrictions are that
I have been running small non-raid
Actually, it is ready in my experience.
/proc/mdstat will tell you how long it will take to fully integrate the new drive.
I raid5 5 external scsi drives. When I want to swap on out, unplug it, plug in the new one, add it to the raid set and your done. No downtime. Simple. Checking
A better option might be to have this taken care of in hardware like bad sectors on some hard drives. Just have the chip do self tests on itself in the spare cycles when it isn't being used.
This would probably be a good option for small, low performance chips or RAM, but probably not for anything like a CPU.
Adding redundancy for high performance chips would require either duplicate cores, one of which would be turned off, or increasing the size of a single core. Increasing the size of the core, however would lead to lower clock speeds and lower performance to let impulses propagate over the extra space.
I would guess, though that turning one of two cores off if it fails a test and selling the cpu as a single core chip will be standard practice when dual core chips go into mass production, much the same way that chips that fail at higher clock speeds are sold as slower chips.
Before you say it, I know some companies mass produce dual core chips now, but I'm thinking mass as in x86 scale, not Power scale.
I assume SMTP settings are separate because SMTP servers are separate from pop/imap/etc accounts. Even though they happen to usually be on the same machine, they provide separate services.
You set the SMTP server to be the one where you want to send email through.
apache2 is not required.
There is a standalone svn server, but if your users have ssh to the server, ssh is probably the easiest option to deal with.
From the administrative point of view, there is another issue that I haven't seen mentioned that svn helps with. The whole thing is kept in a single file, so you don't have many of the problems with group ouwnership that cvs has. Creating a new file in the repository is just a db write, so there is no new file, where cvs would create a new rcs file, which would probably need its group changed before anyone can use it.
1) Give me an example of these problems you're referring to. I really can't think of any way that app bundles are inferior to the Unix way.
/usr can be mounted read-only and shared by many machines over a network, so updates only have to be done once. To do updates, the drive is remounted read-write temporarily and then reverted to read-only. Being read-only brings both performance and security gains. Sharing saves disk space on clients and centralizes administration. /etc can (usually) be mounted read-only, except when doing config changes, but needs to be specific to a machine. /var needs to be read/write and specific to a machine. /home needs to be read/write, but may be shared.
The most glaringly obvious example is that the current structure is made so that as many files as possible can be 1)mounted on read-only drives 2)shared across machines.
If you read the fhs, it makes these points very clear.
2) What's wrong with giving the *users* of the system the easiest way of doing things and letting the Administrators or Developers, the people who KNOW computers, doing the troubleshooting? The users can't troubleshoot; Administrators and Developers can.
Right now, users are more than welcome to keep packages in their own directory if they want. They can use any system they want and either set their path or link from ~/bin.
The question I would have for someone proposing the each package is a self contained heirarchy solution is "What are the advantages of using these packages as opposed to using a good package management system?" With a good package manager, if you want to know what files are part of a pachage, ask the package manager. If you want to remove the package, ask the package manager to. If you want to know what is installed, ask the package manager.
I was thinking the same thing. Think about the cost of console games. You could buy an xbox or a ps2 and have $35 left over, or for another $10 you could get 2 game cubes.
That doesn't really incorporate the whole problem either.
Big Iron also wins for applications that parallelize well, but require lots of communication amoung the nodes.
The reality however is that the best tool for the jobs is highly dependant on the individual application and it's hard to give a vague classification of what will work best in what situation.
The other issues are not just value and what works best, but what gives the best value. A cluster may provide the same performance for less money, but require an extra 100 man-days to code the application, while a big iron server may cost more , but save you the 100 man-days of work. Additionally, there will usually be less work to be done replacing pieces of the system simpley because there are a smaller number of pieces to break and they are usually higher quality.
We (University of Kentucky) just moved away from having the ssn on student id's, but it is still the same as our student identification number (for registration, billing, etc).
It's called "Representative Democracy".
I would also ask what president has lost the popular vote beyond the margin of error?
The electoral college is in place and had lasted this long for a reason. It forces candidates to go after voters across the nation/gives all states a say in the election.
The college has the same effect as the congress. Smaller states are given the same number of senators so that they are represented nationally. Larger states are given more representatives because they have more people and pay more taxes. States are given electoral votes so that smaller states have a say.
Without the electoral college, candidates would simply pander to NY, LA, Chicago, and maybe a few other markets. They can then safely ignore the rest of the nation because a small victory in the large markets overcomes even the largest losses in the rest of the nation.
The other reason for using the electoral college is that it creates a consensus. This year was a big enough victory that it doesn't really matter, but four years ago, the popular vote was too close to call(Let it go.), but the electoral college presented a clear winner(eventually). I think that after some thought, people on both sides of the issue would agree that if the election is really that close, we are better off with either candidate as president than with the presidency vacant.
I agree that clarity comes in handly, but I think the main issues you may have are with the distributors.
I use Fedora (bring on the flames) at home and a Mac at work. Everything I can think of on the Mac is easy to find on my Fedora box. Whatever you have chosen for email is labeled email. web is "Firefox Web Browser". Text editor is "Text Editor".
As for programs like vi, gawk, sed, etc. anyone whoe needs to use or knows how to use any of these as well as grep, lex, bison, emacs knows the commands right off.
With the latest versions of tcp, fast recovery results in behavior as described.
Actually, TCP increases exponentially until the first packet is dropped. It backs off to half, then increases linearly, until another packet is dropped, backs off to half ...
This means that a TCP connection uses ~75% of the bandwidth available to it (after all this stabilizes). So if there is only a single tcp connection over a given length, it will be 75% full at best. However, the whole reason for doing a lot of this is to allow many connections to coexist. If you transmit as fast as possible, you will get the highest throughput possible, but you will end up with a lot of dropped packets and won't play nice with others.