It locks them in by (a) creating a closed network, such that when using skype, you can only communicate with other skype VoIP users (note that I'm talking about only across the Internet, not using the PSTN gateway), and (b) if skype is successful enough, the costs of the skype user-base moving to another standard (open or not) becomes so high that it doesn't happen. The skype user-base are then locked in. Microsoft's MS Word monopoly for a classic example of what happens when users are locked into a proprietory format for their information.
For example, I don't use MS Word, and, theoretically, I can't apply for a number of jobs I see, because they want my resume in MS Word format. Luckly for me, OO.org have reverse engineered the MS Word format, so I can convert my resume. However, if Microsoft introduce any sort of copy protection / DRM / encryption into the MS Word format (and I think they are planning to), then reverse engineering of the MS Word format by the OO.org group would be illegal under the DMCA. If that happens, then either I can't apply for a job I may be a perfect candidate for, or if I want to apply, I then have to spend money on an OS I'd prefer not to run (I run Linux exclusively these days), and by a Word processor I don't really need (OO.org is fine for what I do). Of course I could go down to Kinkos or something, which would be a cheaper alternative. However, the cheapest alternative would be to use the software I already have, on a computer I already have. Yet I couldn't, because the people I'm sending my information to are locked into MS Word, and that forces me to try to adapt, increasing my costs. The only people who win out of a scheme like this is Microsoft.
Now, imagine if the people advertising jobs asked for resumes in open formats such as HTML or PDF format. I could choose tools that I like to generate them (I might like to use vi for HTML or postscript, and then convert the postscript to PDF), and other people can choose to use different tools that suit the way they work.
All the "standard" SIP soft phones that I have tried have been so complicated that I've decided to keep this unstandard little piece of software until something better emerges.
At least under Linux, I've found Linphone to be a very simple and easy to use soft implementation of a SIP phone.
So if I wanted to make an Internet-only call, are you suggesting I should have my call hop off the Internet onto the PSTN (open standard VoIP -> open standard PSTN), then from the PSTN back on to the Internet (open standard PSTN -> proprietory Skype), just to talk from an open standard Internet VoIP end point ? Do you really expect me to throw away the cost benefits of Internet telephony just to support Skype's proprietory VoIP ? Have you bought into the Skype technology so much such that you are willing to ignore the major advantage of VoIP, that being reduced call costs ? I think that's what the definition of a "zealot" might be - somebody who believes in something so much that they are willing to ignore the reality of why something exists in the first place.
By the way, I'm the only supplier of Sony batteries in the world for your Sony goods, and because I'm the sole supplier, I've decided to sell them for $40 each. They might look like AAs, but you can only use Sony ones in your Sony devices, because the Sony ones have a special Sony-only interface (I'm picking on Sony because their memory stick technolgy is proprietory, yet an open standard equivalent exists). What sort of car do you drive, as I may be the sole, world-wide provider of petrol / gasoline for it as well.
You can blame it on whomever you want, the fact remains that the Internet is a very unreliable network by the time you get to end users, and the users are not going to stand for it.
If you are so unhappy with it, why are you still using it ?
'them young whipper-snappers.'
My comments about ego being greater than care factor are independent of age.
or would you prefer to only be able to use Sony batteries in your Sony devices, Ford petrol (gasoline) in your Ford car, etc.
I'm certainly pro-open source, however, I think open standards are far, far more important than the open / closed source issue. If the best tool for the job is closed source, that's fine. I use closed source Adobe Acrobat under Linux rather than xpdf, as I find it better and quicker to use. I would abandon Adobe Acrobat if Adobe closed the PDF specification though, as the open specs of PDF are more important to me than the Adobe Acrobat software itself.
but in the last year, I have probably had almost a month of total downtime, twice the length of downtime was almost a week. I don't live in the boondocks either, my ISP, a national cable provider here is based in the city I live in.
I don't think that is an inherent limitation of the technology, I think that is a limitation of the people deploying the technology. In my experience, a higher than acceptable number of network engineers don't treat the network as carefully as they should. They take somewhat of a "maverick" approach - their ego regarding their level of skills exceeds their care for continued service availability for their customers.
Cisco abandoned a slogan a while back which I thought was great, and is my philosophy on how a network should be designed and run - "The network works, no excuses." The goal should be that faults in a network should never be attributable to a human, if they are, that indicates a lack of planning and care in managing and operating the network.
Sadly, in reality, humans are the greatest cause of network faults. According to RFC 3439 - Some Internet Architectural Guidelines and Philosophy, "... 80% of unscheduled outages are caused by people or process errors [SCOTT]". The best way to run a network is to only "touch" it when you absolutely have to, and question whether that "touching" is necessary twice - much like the old carpenter's saying, "measure twice, cut once".
I figured as much, however, I think this issue really is best judged from the individual's point of view, as they are the ones "suffering" from the emergency. In the past, they were only had the option of relying on a single phone line, or using their car. Now there are multiple, alternative communications channels available to them in an emergency.
and if you're referring to using UDP over tunnels to get around the problems NAT causes, look up "NAT traversal" in google. It is being introduced to both IPsec and SIP.
Skype doesn't do anything special, other than lock its "customers" into a proprietory VoIP network.
This "911" / emergency call issue is regularly rolled out as a "problem" with VoIP, due to the concern about the perceived lack of reliablity of IP based networks. While it is legitimate to address it, I don't think it is the issue it is made out to be.
If you consider what life was like 20 years ago, each house only had one land line, and all the land lines in the neighbourhood were attached to the same exchange. From the individual end user's point of view, there was no redundancy at all. If your land line failed, you couldn't go next door to use your neighbour's phone in an emergency, as their's was dead too. All you could do was jump in your car and go to the hospital, police station or firehouse.
Today, not only do we still have traditional land lines, we also have cell / mobile phones, and we still have cars. If we add VoIP into the mix, in sum, we have significantly more available "emergency communcations" resources as individuals than we did 20 years ago. That's why I don't think "911" / emergency call services is the issue it is made out to be when thinking about deploying VoIP.
Make sure cwd is the last entry in the path. In fact, it is actually better not to have it in there at all. Here is why.
If you put cwd as the first entry in the path, of course the shell will always look in the cwd for the program requested. This can be convenient if the user happens to name their program "ls". The shell will run "local" cwd versions of any and all system binaries.
However, if an attacker executes a bit of social engineering, they can get the user to run a program they don't intend to. That program can do some "interesting" things, like change the users password on them, without them knowing or realising.
Here is an scenario. The attacker walks up to a student not aware of social engineering tricks like this one. "Hi, can you check something for me, I've left a file called foo in/tmp, and I'd like to see if it is still there." The user goes to/tmp, and executes "ls foo", expecting/bin/ls to be executed. However, the attacker has placed an "ls" binary in/tmp, which (a) changes the users password silently to something the attacker knows, and (b) executes the real/bin/ls, giving the appearance that the only thing executed was/bin/ls.
Of course, the student's account has now been compromised. Can you imagine what would happen if that "student" was the root user in the above scenario, and had cwd in the path ?
When IBM designed the original PC, they used a 8086 (or 8088, I can't exactly remember) CPU, which supported an address space of up to 20 bits. 20 bits allows addressing of up to 1MB of RAM.
The original IBM PC was quite limited - the only storage IO device the BIOS supported was a audio cassette interface. The original BIOS didn't support FDDs or HDDs. To allow for these and other devices to be added, IBM allowed for BIOS extensions to be added, with the installation of more BIOS chips or extension cards in the ISA slots. BIOS extensions have to appear somewhere within the address space of the CPU, yet not within the same region as the system RAM. So they decided on placing that boundary between RAM and ROM Extensions at the 640KB mark.
384KB of address space (1024 - 640) for additional devices probably sounds, and probably ended up being excessive. However, bare in mind that the original IBM PC only came with 16KB of RAM. The size of both the supported RAM, and address space allocated for extention devices, probably sounded so large at the time that these limits were probably assumed to be unreachable. History has shown otherwise, probably because the IBM PC architecture was a success beyond anybody's predictions.
Bill Gates may have made the statement that "640 KB is enough for anybody". If he did, then he would have only been endorsing the 640KB barrier that IBM had decided upon, rather than being responsible for it.
up until the point where you win ? If Microsoft chose to drag it out, they could easily bankrupt you, forcing you into a position of accepting paying a fine. Morally and legally you'd be right, the question then becomes have you enough money to prove that.
I don't think I did. I was mearly suggesting ways to address your issue with validating binaries.
I live in Australia. We only use paper ballots to count around 12 to 15 million votes (I can't remember the exact number of eligable voters) - we did only three weeks ago. I see no problem with it. It works for us, I don't know what the problem is in America.
Then again, we separate local, state and federal government, and therefore have separate local, state and federal elections. Voting in local elections is optional, state and federal isn't. From what I understand, one of the issues that makes electronic voting attractive in America is the loading up of state related questions on a "federal" ballot. I seem to remember hearing that some Americans needed 10 minutes to vote, just to work through all the state related issues.
When I voted recently in our Federal election, it took about a minute, with a couple of sheets of paper and a pencil.
That being said, electronic voting, with auditable procedures and software, plus a voter auditable paper trail, would be a possible optimisation, but not a complete replacement for manual counting of votes.
I think the broad issue here isn't that electronic voting can't be made to work in an auditable manner. I think the real issue here is why is there resistance to making it properly amd thoroughly independently auditable, via the addition of a paper trail ?
The reality is that electronic records of the vote require the humans trust the machine. Open Source or closed, the binaries on the machine can not be directly examined, rendering the nature of the code used a moot point.
I think you can perform this sort of verification, very easily, much the same as you verify an ISO file that you download from the net.
At the start of the election process, all binaries are verified using MD5 or SHA1, or one of the more resent (and stronger) hashes. A trusted third party performs the hash. Note that the result isn't secret information, and in fact should be widely published allowing multiple people to verify the results independently if necessary. The source/binaries could be downloaded from the net, the results of the official hash could be a public event, and reported on the news, allowing interested voters to independently perform their own validation.
After the election process, the hashes can be re-verified, again both publically and by a trusted party, a judge perhaps, to ensure the binaries weren't modified.
Of course, the binary for the hashing program could be modified, however you could require open source code for that published and available, and have the development / compilation process public and auditable.
Mind you, you could still be vulnerable to OS or hardware level attacks eg. CPU hardware modified, presuming one of the designers of the CPU at, say, Intel was corruptable. These again can be remedied / mitigated by ensuring a public and auditable process. One way to mitigate the CPU level attack would be to run the software on the multiple of different x86 vendor CPUs eg AMD, Intel, VIA and Transmeta.
If you follow this process through, you actually start to realise that absolute security is a fallacy. However, by involving a number of different people, with different trust levels and expertise, making the process public and auditable, using independant hardware and software, you can increase significantly the ammount of assurance that the result is far more trustworthy.
I can just imagine telling someone "yeah, here's a free operating system for ya, you just have to pay to run anything you need to run" LOL!
I'm curious to know the extent of your "virtualisation" of Win98, in the sense that is your goal to allow the user to run all their previous Win98 applications ?
The reason I ask is that if it is, then, as much as I'm completely pro-Linux (I don't run anything else), I'd wonder if you are doing your end-users a disservice. If you going down that path, then Linux just becomes a somewhat unnecessary overhead, as I can't see how the users will transition to Linux when all their applications are running under Windows 98, even though it is on a virtualised box.
I think there are two important criteria for a transition to Linux to be successful.
Firstly, around 50% or more of the applications the end-user is going to run need to be native Linux applications. If they aren't, then I'd think there is still too much "gravitational mass" pulling the user back towards Windows.
Secondly, moving between the native and non-native applications, and moving or switching between the native and non-native data needs to be simple for the end-user. Again, if it isn't, the end-user will probably just want to stick with what they know, the non-Linux solution.
Thanks for you comments, I appreciate it, and I do agree with the principle.
Based on the few things I've heard and seen regarding the way Theo deals with people on certain issues, I would wonder whether if a "formal" invitation had been put the LKML, and the Linux community jumped up to support it, Theo would have then complained, possibly in a public forum such as a news paper, that the "Linux crowed took over". His motives can sometimes seem to be not completely altruistic.
While the issue I'm commenting is, at the end of the day, fairly petty, I think it would be useful for Theo to realise that criticism such as his can only hurt his cause, not help it. It is better to promote the people who help you, when they do, and forget about it when they don't; it will only encourage more to offer their help, when they see they get recognised for their contributions.
I don't buy the argument that its the extra registers, because there have been over 56 registers available for register renaming since the early-mid 90's.
I'm no expert, however, from what I understand from the bit if reading I've done and the bit of assembler I've done, it isn't the number of registers on the chip, it is the number of registers available to the user of the chip.
For example, on the classic 32 bit X86, there are only four general purpose registers - EAX, EBX, ECX and EDX. If you want to multiply four numbers together, you can only hold three of them on the CPU, as you need one of those registers to store the result of the multiplication for each round. After three multiplies, you'll have to move the fourth number from main memory into a register, and then perform the fourth multiply.
On the AMD64, there are 12 general purpose registers (I think), so you could store all the digits on the CPU while the multiplication is taking place. RAM is very slow verses the CPU registers, avoiding getting data from RAM during the calculation is worth while, as the CPU isn't delayed, waiting for data from RAM, part way through doing the calculation.
Of course, for the above examples, avoiding going to RAM for the forth digit is probably not going to make a significant or measurable difference. However, imagine if that calculation was being performed a 1 000 000 or a 100 000 000 times - the small saving adds up when that saving occurs many, many times, making it a big saving, which can result in a significant performance increase.
That is why more, exposed general purpose registers on a CPU are useful.
Now. If Nvidia open sourced its drivers you know what would happen? ATI would look at the code and learn how to make better drivers. Then the ATI driver would be just as good if not better than the NVidia. Linux users wouldn't choose nvidia because of better support because ATI would be a reasonable alternative.
I disagree with this. There are plenty of open source drivers in the Linux kernel for high performance hardware that would provide examples for ATI to follow.
One example would be the Intel, GPL licensed driver for the Intel 10 GBps Ethernet cards. Another example would be the GPL open source drivers for the s2io XFrame 10 Gbps Ethernet cards. I'm sure there would be others, probably part of the IBM mainframe support open source code.
I think he has earned the right be hostile if he wants to- it works.
From the Age article:
He said he found it sad that the Linux crowd did not help in the activism at all. "(They) always seem to talk about freedom but are not helping in this activism. It's basically BSD people doing it. That is curious. For instance, do you think Linus (Torvalds) would send a mail to TI? No, I would bet money that he did not. Yet he is aware of what is going on. That's very odd to me."
Do you think this will go across well with the "Linux crowd". "Betting money" that Linus wouldn't help ? Questioning Linus's commitment to freedom drivers? Do you think Theo has made more or less friends with this statement ?
Where was his announcement to the LKML (which is probably the most central contact point for the "Linux crowd") that he was going to pursue the vendors, and an invitation for the Linux community to help with that effort ? Did he email Linus directly ?
Looking at LKML briefly, the only post I can find specific to this effort is here, dated the 28th of October. It suggests that the effort has already been on going.
In other words, it appears that the "Linux crowd" didn't know this was happening until after it had started.
It seems that Theo has forgotten that he didn't invite the "Linux crowd" to participate in the first place.
I don't want to start a licensing flamewar (it's my opinion, I'm allowed to have one, and you're allowed to have a different one), however, in my opinion Linus is more supportive of openness than Theo is, because the GPL requires continued ancestral openness, where as the BSD license doesn't.
I wonder if Linus could do something similar to get ATI and NVidia to open up...
The day he picked the GPL, and the day he made a conscious decision to enforce it with the arrival of binary modules was the day he did all he could. On many occasions he (and others) have provided good technical justification for their decision to only support open drivers. If the vendors such as ATI and NVidia don't agree with or see the benefits of this approach, there isn't much Linus can do.
As you seem to have missed it, the issue is the behaviour of TCP when it operates over a path involving links which have asymmetric bit per second values.
If you run smartmontools, you can configure smartd to not only monitor the SMART status of the disks, but also execute online tests - have a look at the "-s" option of smartd. For my RAID1 array, for each device, I have -s (L/../../7/03|S/../.././05) entries.
mdadm also has a daemon mode which can monitor the arrays, and if there are any failures, send an email to a designated email address.
Slashdot Mirror
How does it lock its customers?
It locks them in by (a) creating a closed network, such that when using skype, you can only communicate with other skype VoIP users (note that I'm talking about only across the Internet, not using the PSTN gateway), and (b) if skype is successful enough, the costs of the skype user-base moving to another standard (open or not) becomes so high that it doesn't happen. The skype user-base are then locked in. Microsoft's MS Word monopoly for a classic example of what happens when users are locked into a proprietory format for their information.
For example, I don't use MS Word, and, theoretically, I can't apply for a number of jobs I see, because they want my resume in MS Word format. Luckly for me, OO.org have reverse engineered the MS Word format, so I can convert my resume. However, if Microsoft introduce any sort of copy protection / DRM / encryption into the MS Word format (and I think they are planning to), then reverse engineering of the MS Word format by the OO.org group would be illegal under the DMCA. If that happens, then either I can't apply for a job I may be a perfect candidate for, or if I want to apply, I then have to spend money on an OS I'd prefer not to run (I run Linux exclusively these days), and by a Word processor I don't really need (OO.org is fine for what I do). Of course I could go down to Kinkos or something, which would be a cheaper alternative. However, the cheapest alternative would be to use the software I already have, on a computer I already have. Yet I couldn't, because the people I'm sending my information to are locked into MS Word, and that forces me to try to adapt, increasing my costs. The only people who win out of a scheme like this is Microsoft.
Now, imagine if the people advertising jobs asked for resumes in open formats such as HTML or PDF format. I could choose tools that I like to generate them (I might like to use vi for HTML or postscript, and then convert the postscript to PDF), and other people can choose to use different tools that suit the way they work.
All the "standard" SIP soft phones that I have tried have been so complicated that I've decided to keep this unstandard little piece of software until something better emerges.
At least under Linux, I've found Linphone to be a very simple and easy to use soft implementation of a SIP phone.
PSTN via SkypeOut.
So if I wanted to make an Internet-only call, are you suggesting I should have my call hop off the Internet onto the PSTN (open standard VoIP -> open standard PSTN), then from the PSTN back on to the Internet (open standard PSTN -> proprietory Skype), just to talk from an open standard Internet VoIP end point ? Do you really expect me to throw away the cost benefits of Internet telephony just to support Skype's proprietory VoIP ? Have you bought into the Skype technology so much such that you are willing to ignore the major advantage of VoIP, that being reduced call costs ? I think that's what the definition of a "zealot" might be - somebody who believes in something so much that they are willing to ignore the reality of why something exists in the first place.
By the way, I'm the only supplier of Sony batteries in the world for your Sony goods, and because I'm the sole supplier, I've decided to sell them for $40 each. They might look like AAs, but you can only use Sony ones in your Sony devices, because the Sony ones have a special Sony-only interface (I'm picking on Sony because their memory stick technolgy is proprietory, yet an open standard equivalent exists). What sort of car do you drive, as I may be the sole, world-wide provider of petrol / gasoline for it as well.
You can blame it on whomever you want, the fact remains that the Internet is a very unreliable network by the time you get to end users, and the users are not going to stand for it.
If you are so unhappy with it, why are you still using it ?
'them young whipper-snappers.'
My comments about ego being greater than care factor are independent of age.
or would you prefer to only be able to use Sony batteries in your Sony devices, Ford petrol (gasoline) in your Ford car, etc.
I'm certainly pro-open source, however, I think open standards are far, far more important than the open / closed source issue. If the best tool for the job is closed source, that's fine. I use closed source Adobe Acrobat under Linux rather than xpdf, as I find it better and quicker to use. I would abandon Adobe Acrobat if Adobe closed the PDF specification though, as the open specs of PDF are more important to me than the Adobe Acrobat software itself.
but in the last year, I have probably had almost a month of total downtime, twice the length of downtime was almost a week. I don't live in the boondocks either, my ISP, a national cable provider here is based in the city I live in.
I don't think that is an inherent limitation of the technology, I think that is a limitation of the people deploying the technology. In my experience, a higher than acceptable number of network engineers don't treat the network as carefully as they should. They take somewhat of a "maverick" approach - their ego regarding their level of skills exceeds their care for continued service availability for their customers.
Cisco abandoned a slogan a while back which I thought was great, and is my philosophy on how a network should be designed and run - "The network works, no excuses." The goal should be that faults in a network should never be attributable to a human, if they are, that indicates a lack of planning and care in managing and operating the network.
Sadly, in reality, humans are the greatest cause of network faults. According to RFC 3439 - Some Internet Architectural Guidelines and Philosophy, "... 80% of unscheduled outages are caused by people or process errors [SCOTT]". The best way to run a network is to only "touch" it when you absolutely have to, and question whether that "touching" is necessary twice - much like the old carpenter's saying, "measure twice, cut once".
I figured as much, however, I think this issue really is best judged from the individual's point of view, as they are the ones "suffering" from the emergency. In the past, they were only had the option of relying on a single phone line, or using their car. Now there are multiple, alternative communications channels available to them in an emergency.
and if you're referring to using UDP over tunnels to get around the problems NAT causes, look up "NAT traversal" in google. It is being introduced to both IPsec and SIP.
Skype doesn't do anything special, other than lock its "customers" into a proprietory VoIP network.
I think skype will kill itself.
This "911" / emergency call issue is regularly rolled out as a "problem" with VoIP, due to the concern about the perceived lack of reliablity of IP based networks. While it is legitimate to address it, I don't think it is the issue it is made out to be.
If you consider what life was like 20 years ago, each house only had one land line, and all the land lines in the neighbourhood were attached to the same exchange. From the individual end user's point of view, there was no redundancy at all. If your land line failed, you couldn't go next door to use your neighbour's phone in an emergency, as their's was dead too. All you could do was jump in your car and go to the hospital, police station or firehouse.
Today, not only do we still have traditional land lines, we also have cell / mobile phones, and we still have cars. If we add VoIP into the mix, in sum, we have significantly more available "emergency communcations" resources as individuals than we did 20 years ago. That's why I don't think "911" / emergency call services is the issue it is made out to be when thinking about deploying VoIP.
cause I'm in trouble now !
Make sure cwd is the last entry in the path. In fact, it is actually better not to have it in there at all. Here is why.
If you put cwd as the first entry in the path, of course the shell will always look in the cwd for the program requested. This can be convenient if the user happens to name their program "ls". The shell will run "local" cwd versions of any and all system binaries.
However, if an attacker executes a bit of social engineering, they can get the user to run a program they don't intend to. That program can do some "interesting" things, like change the users password on them, without them knowing or realising.
Here is an scenario. The attacker walks up to a student not aware of social engineering tricks like this one. "Hi, can you check something for me, I've left a file called foo in /tmp, and I'd like to see if it is still there." The user goes to /tmp, and executes "ls foo", expecting /bin/ls to be executed. However, the attacker has placed an "ls" binary in /tmp, which (a) changes the users password silently to something the attacker knows, and (b) executes the real /bin/ls, giving the appearance that the only thing executed was /bin/ls.
Of course, the student's account has now been compromised. Can you imagine what would happen if that "student" was the root user in the above scenario, and had cwd in the path ?
When IBM designed the original PC, they used a 8086 (or 8088, I can't exactly remember) CPU, which supported an address space of up to 20 bits. 20 bits allows addressing of up to 1MB of RAM.
The original IBM PC was quite limited - the only storage IO device the BIOS supported was a audio cassette interface. The original BIOS didn't support FDDs or HDDs. To allow for these and other devices to be added, IBM allowed for BIOS extensions to be added, with the installation of more BIOS chips or extension cards in the ISA slots. BIOS extensions have to appear somewhere within the address space of the CPU, yet not within the same region as the system RAM. So they decided on placing that boundary between RAM and ROM Extensions at the 640KB mark.
384KB of address space (1024 - 640) for additional devices probably sounds, and probably ended up being excessive. However, bare in mind that the original IBM PC only came with 16KB of RAM. The size of both the supported RAM, and address space allocated for extention devices, probably sounded so large at the time that these limits were probably assumed to be unreachable. History has shown otherwise, probably because the IBM PC architecture was a success beyond anybody's predictions.
Bill Gates may have made the statement that "640 KB is enough for anybody". If he did, then he would have only been endorsing the 640KB barrier that IBM had decided upon, rather than being responsible for it.
up until the point where you win ? If Microsoft chose to drag it out, they could easily bankrupt you, forcing you into a position of accepting paying a fine. Morally and legally you'd be right, the question then becomes have you enough money to prove that.
I don't think I did. I was mearly suggesting ways to address your issue with validating binaries.
I live in Australia. We only use paper ballots to count around 12 to 15 million votes (I can't remember the exact number of eligable voters) - we did only three weeks ago. I see no problem with it. It works for us, I don't know what the problem is in America.
Then again, we separate local, state and federal government, and therefore have separate local, state and federal elections. Voting in local elections is optional, state and federal isn't. From what I understand, one of the issues that makes electronic voting attractive in America is the loading up of state related questions on a "federal" ballot. I seem to remember hearing that some Americans needed 10 minutes to vote, just to work through all the state related issues.
When I voted recently in our Federal election, it took about a minute, with a couple of sheets of paper and a pencil.
That being said, electronic voting, with auditable procedures and software, plus a voter auditable paper trail, would be a possible optimisation, but not a complete replacement for manual counting of votes.
I think the broad issue here isn't that electronic voting can't be made to work in an auditable manner. I think the real issue here is why is there resistance to making it properly amd thoroughly independently auditable, via the addition of a paper trail ?
The reality is that electronic records of the vote require the humans trust the machine. Open Source or closed, the binaries on the machine can not be directly examined, rendering the nature of the code used a moot point.
I think you can perform this sort of verification, very easily, much the same as you verify an ISO file that you download from the net.
At the start of the election process, all binaries are verified using MD5 or SHA1, or one of the more resent (and stronger) hashes. A trusted third party performs the hash. Note that the result isn't secret information, and in fact should be widely published allowing multiple people to verify the results independently if necessary. The source/binaries could be downloaded from the net, the results of the official hash could be a public event, and reported on the news, allowing interested voters to independently perform their own validation.
After the election process, the hashes can be re-verified, again both publically and by a trusted party, a judge perhaps, to ensure the binaries weren't modified.
Of course, the binary for the hashing program could be modified, however you could require open source code for that published and available, and have the development / compilation process public and auditable.
Mind you, you could still be vulnerable to OS or hardware level attacks eg. CPU hardware modified, presuming one of the designers of the CPU at, say, Intel was corruptable. These again can be remedied / mitigated by ensuring a public and auditable process. One way to mitigate the CPU level attack would be to run the software on the multiple of different x86 vendor CPUs eg AMD, Intel, VIA and Transmeta.
If you follow this process through, you actually start to realise that absolute security is a fallacy. However, by involving a number of different people, with different trust levels and expertise, making the process public and auditable, using independant hardware and software, you can increase significantly the ammount of assurance that the result is far more trustworthy.
I can just imagine telling someone "yeah, here's a free operating system for ya, you just have to pay to run anything you need to run" LOL!
I'm curious to know the extent of your "virtualisation" of Win98, in the sense that is your goal to allow the user to run all their previous Win98 applications ?
The reason I ask is that if it is, then, as much as I'm completely pro-Linux (I don't run anything else), I'd wonder if you are doing your end-users a disservice. If you going down that path, then Linux just becomes a somewhat unnecessary overhead, as I can't see how the users will transition to Linux when all their applications are running under Windows 98, even though it is on a virtualised box.
I think there are two important criteria for a transition to Linux to be successful.
Firstly, around 50% or more of the applications the end-user is going to run need to be native Linux applications. If they aren't, then I'd think there is still too much "gravitational mass" pulling the user back towards Windows.
Secondly, moving between the native and non-native applications, and moving or switching between the native and non-native data needs to be simple for the end-user. Again, if it isn't, the end-user will probably just want to stick with what they know, the non-Linux solution.
I musn't be a mainstream Linux user anymore by the looks of it, because I don't want to run a virtualised version of any MS products.
I think mainstream Linux is still in the server space, and I think Xen (and qemu and UML) in those environments would be very useful.
Heh, you've hit the edge of my assembler knowledge, and I didn't think the example through that well..
However, the point I was trying to show was that on a processor with additional GP registers, you would be able to add to your example
If such an "eex" register existed, instead of
In other words, the additional GP registers allow both the number of "mov" instructions, and the delays they cause, to be reduced,.
Thanks for you comments, I appreciate it, and I do agree with the principle.
Based on the few things I've heard and seen regarding the way Theo deals with people on certain issues, I would wonder whether if a "formal" invitation had been put the LKML, and the Linux community jumped up to support it, Theo would have then complained, possibly in a public forum such as a news paper, that the "Linux crowed took over". His motives can sometimes seem to be not completely altruistic.
While the issue I'm commenting is, at the end of the day, fairly petty, I think it would be useful for Theo to realise that criticism such as his can only hurt his cause, not help it. It is better to promote the people who help you, when they do, and forget about it when they don't; it will only encourage more to offer their help, when they see they get recognised for their contributions.
See my earlier post as to why.
I don't buy the argument that its the extra registers, because there have been over 56 registers available for register renaming since the early-mid 90's.
I'm no expert, however, from what I understand from the bit if reading I've done and the bit of assembler I've done, it isn't the number of registers on the chip, it is the number of registers available to the user of the chip.
For example, on the classic 32 bit X86, there are only four general purpose registers - EAX, EBX, ECX and EDX. If you want to multiply four numbers together, you can only hold three of them on the CPU, as you need one of those registers to store the result of the multiplication for each round. After three multiplies, you'll have to move the fourth number from main memory into a register, and then perform the fourth multiply.
On the AMD64, there are 12 general purpose registers (I think), so you could store all the digits on the CPU while the multiplication is taking place. RAM is very slow verses the CPU registers, avoiding getting data from RAM during the calculation is worth while, as the CPU isn't delayed, waiting for data from RAM, part way through doing the calculation.
Of course, for the above examples, avoiding going to RAM for the forth digit is probably not going to make a significant or measurable difference. However, imagine if that calculation was being performed a 1 000 000 or a 100 000 000 times - the small saving adds up when that saving occurs many, many times, making it a big saving, which can result in a significant performance increase.
That is why more, exposed general purpose registers on a CPU are useful.
Now. If Nvidia open sourced its drivers you know what would happen? ATI would look at the code and learn how to make better drivers. Then the ATI driver would be just as good if not better than the NVidia. Linux users wouldn't choose nvidia because of better support because ATI would be a reasonable alternative.
I disagree with this. There are plenty of open source drivers in the Linux kernel for high performance hardware that would provide examples for ATI to follow.
One example would be the Intel, GPL licensed driver for the Intel 10 GBps Ethernet cards. Another example would be the GPL open source drivers for the s2io XFrame 10 Gbps Ethernet cards. I'm sure there would be others, probably part of the IBM mainframe support open source code.
I think he has earned the right be hostile if he wants to- it works.
From the Age article :
He said he found it sad that the Linux crowd did not help in the activism at all. "(They) always seem to talk about freedom but are not helping in this activism. It's basically BSD people doing it. That is curious. For instance, do you think Linus (Torvalds) would send a mail to TI? No, I would bet money that he did not. Yet he is aware of what is going on. That's very odd to me."
Do you think this will go across well with the "Linux crowd". "Betting money" that Linus wouldn't help ? Questioning Linus's commitment to freedom drivers? Do you think Theo has made more or less friends with this statement ?
Where was his announcement to the LKML (which is probably the most central contact point for the "Linux crowd") that he was going to pursue the vendors, and an invitation for the Linux community to help with that effort ? Did he email Linus directly ?
Looking at LKML briefly, the only post I can find specific to this effort is here, dated the 28th of October. It suggests that the effort has already been on going.
In other words, it appears that the "Linux crowd" didn't know this was happening until after it had started.
It seems that Theo has forgotten that he didn't invite the "Linux crowd" to participate in the first place.
I don't want to start a licensing flamewar (it's my opinion, I'm allowed to have one, and you're allowed to have a different one), however, in my opinion Linus is more supportive of openness than Theo is, because the GPL requires continued ancestral openness, where as the BSD license doesn't.
I wonder if Linus could do something similar to get ATI and NVidia to open up...
The day he picked the GPL, and the day he made a conscious decision to enforce it with the arrival of binary modules was the day he did all he could. On many occasions he (and others) have provided good technical justification for their decision to only support open drivers. If the vendors such as ATI and NVidia don't agree with or see the benefits of this approach, there isn't much Linus can do.
As you seem to have missed it, the issue is the behaviour of TCP when it operates over a path involving links which have asymmetric bit per second values.
I suggest reading RFC 3449 - TCP Performance Implications of Network Path Asymmetry for further information.
There is a pretty good chance he knows what he is talking about ...
If you run smartmontools, you can configure smartd to not only monitor the SMART status of the disks, but also execute online tests - have a look at the "-s" option of smartd. For my RAID1 array, for each device, I have -s (L/../../7/03|S/../.././05) entries.
mdadm also has a daemon mode which can monitor the arrays, and if there are any failures, send an email to a designated email address.