Another problem with "lifetime subscriptions" is that the "lifetime" may be explained as the lifetime of yourself, the lifetime of the box, or the lifetime of the service.
Witnessing the quick end of many services, most customers will be reluctant to pay $500 upfront for "lifetime" service on something like a TiVo, where that service may end next month because of bankrupcy or because the company no longer feels like providing it. When you pay that much money for an appliance, you would want at least some guarantees (like it will continue to work for the lifetime of the box when the company folds next month)
The answer to that is: it may get worse overnight, but it does not get worse and worse from that. The first night it will run updatedb and possibly other scripts that fill the disk cache, but it is not a memory leak or something else that claims more and more memory.
In fact on one server at work I do run a program that claims all memory and frees it, early in the morning after all those jobs have run. This improves things when memory is a bit short.
Sometimes a valid reason can be that the disk can sleep when it is not accessed, and would need to startup (takes time and wears it out) every time a swap page is touched, which occurs randomly.
This is not a very likely scenario in a general purpose PC, but I do have a TV satellite receiver that runs Linux and normally runs without swap. It uses the disk only for video recording, but there is an option to add swapspace on that disk. This does not work well because the disk is almost always asleep (when you are just watching TV) and it is a nuisance to have it startup at random moments, or when you press some non-recording button on the remote...
Re:We really need to repurpose the sticky bit...
on
Is Swap Necessary?
·
· Score: 1
In Linux, the entire kernel is unswappable.
Re:We really need to repurpose the sticky bit...
on
Is Swap Necessary?
·
· Score: 1
This would probably not work as well as it did in the PDP-11 days.
Then, a program was a single file that was always loaded in memory as a whole. Nowadays, the system uses virtual memory (so the program is not loaded all in RAM, but only the parts that are actually used are loaded). And many programs consist of a main program and some or many dynamically loaded libraries. Would you propose that all of these would be loaded into RAM? And would the sticky bit have to be set on each of these, or would the bit on the program itself imply that all its libraries would have to remain in RAM as well?
Don't you run the risk that newbies would set so many sticky bits that there main applications can remain loaded, but the remaining RAM is too small to accommodate the many other processes in the system, forcing these to swap and to cause disk performance problems for the main applications as well?
Some manual management capabilities can be nice, but they require competent administrators. The wide use of computers without any competent administrations has brought the (Internet) world where it is today, in an ever worsening virus and worm disaster. I don't think relying on competent administration is a good plan.
What happens (also on Linux) is that while you are away, some background process runs that reads a lot of disk data. Things like a backup, the deamon for updating the database for locating files based on name and/or content, etc. This background process tempts the system to believe that it needs lots of pages for disk cache, and it better swaps out program memory. After the process is finished, disk pages remain in the cache and the programs remain on swap, until you touch the mouse.
The system should examine the disk cache, the amount of free memory, the amount and age of swapped pages, and the activity of the disk. When the disk is idle and there is a lot of free memory and/or there are a lot of cached disk sectors that are not used recently, it could copy back pages from swap to RAM (maybe leaving them valid in swap as well until they are touched in RAM).
Then, it can prepare for the user to start using the foreground programs again before he touches the mouse.
Re:Reminds me of issues we had from Welchia
on
Is Swap Necessary?
·
· Score: 1
Well, in cases like this (the ARP table or some other network-related table of pending operations using up all available memory) it is just as valid to claim that the OS should limit the amount of memory spent on such things.
This depends on whether the program can do a better job at memory allocation and swapping than the operating system. This may be true is specific situations, e.g. when the program knows it is done with operating on a specific memory page it can write it back and free the memory, whereas the OS will have to learn this by overseeing the situation.
When the addressing space allows (64-bit system) it is possible to build an OS where the entire disk is in fact a virtual address space, of which the RAM is only a "cache". Programs can just access the disk as if it were memory, and the OS will handle all the reading and writing. With the possibility of "hints" from the application, this would probably work quite well for applications operating on lots of data.
This is not completely true. When you add some swap to your systems you will find that there is some margin where it will use the swap but not slowdown. This is because some processes allocate memory that they later (almost) never touch. This will be swapped out, and your RAM will be used for better things.
I don't know how swapping to a file performs these days, but "back in the old days" (when swap was really required because you could not afford the RAM necessary) it was noticably slower than a swap partition.
Just like you make sure that you have enough RAM, I make sure that I allocate a swap partition (on the first cylinders of the HDD) of 1-2 times RAM size. When it is not required, so be it. But my 1GB RAM system has 100MB of swapspace in use, and I have not run any very-large applications in the 41 days it is up.
Apparently it is possible to program a crappy UI on any OS, including Linux. It is also possible to build hardware that can only recover from crashes via a powercycle.
That is hardly news. It is also hardly news that it is possible to build such boxes and use Linux to good effect. TiVO and the Dreambox are examples.
It is a nice idea but it shout *not queue mail*. The reason why individual customers may want to run their own server is to know what happens with their mail. The mail proxy should only monitor the traffic while it relays the packets between the customer and the server he connects. When the outgoing mail stream shows virus signatures, the customer should be blocked from outgoing port 25 traffic and informed about this.
>Given the gigantic expansion of broadband, I'm surprised that cable / dsl modems don't just do NAT and other firewalling techniques by default.
They don't?? Most modems used here do! E.g. The Alcatel SpeedTouch 510, the Draytek Vigor series. They all have built-in NAT and firewalling.
In its default configuration, it helps against Blaster and Sasser, but not against mailviruses and the resulting spam trojans. So outgoing port 25 blocking would need to be configured by the user. The user is too clueless to do that. (if not, he would have installed a virus scanner)
>The only way to selectively allow individual ip addresses to be able to use outbond would be to have individual allow statements for each customer who requested it placed on the ACL.
Of course that is not the *only* way. The first alternative you can think about is to have separate IP ranges of filtered and unfiltered customers, and then put the customer in the correct range depending on his blocking status. Broadband customers normally have no guaranteed fixed IP.
They can simply block all outgoing port 25 traffic from their customers without effect on most of them.
I don't know how Comcast operates their broadband service, but here the providers often have some "service centre" page accessible to the customer, where they can configure things like pop mailbox names, virus scanning, spam filtering, etc. On such a page there could be a selection for port filtering. By default some ports will be blocked for everyone (outgoing 25, incoming 25, 80, 137-139, etc) and the customer can enable them when he thinks that is necessary.
By having the filtering on by default, all the ordinary clueless users who just want to browse and mail via the ISP mailservers will have protection from many attacks. Those who click on the enables probably know how to install a scanner and firewall.
Comcast itself could transparently proxy all web access to a server that outputs that information. I think it is a good idea.
Furtermore, I think that Internet providers should implement a standard method for reporting infected PC's by IP address and timestamp. They can forward this message to their customer.
Of course you will still need the phone lines! Unless they are replaced by fiber to the home, or so. But that costs so much that rarely anyone is doing it.
>In Norway the Storting (Parliament) has decided to spend $0.5 billion to build an OTA infrastructure to transmit digital TV signals. Once the new network is in place, the old analog network will be switched off. And you thought the US digital TV mandate was bad...
>My prediction is that before the new digital TV network is in place it will already by severly outdated, completely overrun by the Internet revolution.
I don't think it will be that bad, unless they take a decade to build it. Digital TV requires broadcast bandwidth of about 300 Mbit/s for a typical channel package, or multicast bandwidth of 10-20 Mbit/s to each and every home. That is something completely different from Voice over IP!!
Such bandwidth is not in place in todays typical Internet access networks, and not everyone has or wants to have such wideband Internet access.
Even now, most providers have no multicast or broadcast capability AT ALL. Before things like TV broadcasting over Internet are even to be considered, the lowlevel network must first be made multicast-capable, so that not every viewer has to setup a private connection to some streaming server.
Worse, the way many ADSL access networks are built, with ATM throughout the network and PPPoA on top of that, even precludes the efficient use of multicast...
DVB-T is a very efficient broadcast method in such an environment, something that will be hard to beat in a wired network with point-to-point characteristics.
>f you invent a method of a frequency multiplier because you need to use one in 4Mhz steps, it is reasonable that your patent would also cover someone using an identical implementation except with 33Mhz steps
That does not sound unreasonable. But when that is the case, *why* does the original patent so explicitly specify the frequency to be 4 MHz?? I was amazed when I read that. I would not expect such an implementation detail to be specified in a patent.
It is an interesting invention to have a PLL-driven frequency multiplier as a clock source for a microprocessor (although such a PLL by itself, and its use to multiply a reference frequency by some factor, is of course a pre-existing invention).
The patent does not say anything about the input and output frequency, but it does mention the 4 MHz intermediate. To me, it seems like an integral part of the claims. Probably one that should not have been included...
It also explicitly claims that the incoming clock is divided down to 4 MHz (this exact figure is give n more than once) and then locked to the new clock.
Intel processors do not use this frequency, they use 33 MHz.
So what is the point of all this? When Intel are violating this patent, they at most are violating some of the claims but not all of them. Isn't that required for it to be a patent violation?
Apparently, major distributors (like SuSE) have also been informed, and now every user can download the update.
Re:Tell us about the relationship between houses..
on
Wiring a Neighborhood?
·
· Score: 1
He is in the Netherlands (Europe). We have some different standards and situations here.
For example, telephone poles are out of the question. The distances are probably small (dense country). Different types of connector are normally used (IEC instead of F). And Radio Shack moved out of here a decade ago.
Slashdot Mirror
Of course the price of the software will go up by the amount the hardware costs. Anyone would understand that.
Another problem with "lifetime subscriptions" is that the "lifetime" may be explained as the lifetime of yourself, the lifetime of the box, or the lifetime of the service.
Witnessing the quick end of many services, most customers will be reluctant to pay $500 upfront for "lifetime" service on something like a TiVo, where that service may end next month because of bankrupcy or because the company no longer feels like providing it.
When you pay that much money for an appliance, you would want at least some guarantees (like it will continue to work for the lifetime of the box when the company folds next month)
The answer to that is: it may get worse overnight, but it does not get worse and worse from that.
The first night it will run updatedb and possibly other scripts that fill the disk cache, but it is not a memory leak or something else that claims more and more memory.
In fact on one server at work I do run a program that claims all memory and frees it, early in the morning after all those jobs have run. This improves things when memory is a bit short.
Sometimes a valid reason can be that the disk can sleep when it is not accessed, and would need to startup (takes time and wears it out) every time a swap page is touched, which occurs randomly.
This is not a very likely scenario in a general purpose PC, but I do have a TV satellite receiver that runs Linux and normally runs without swap. It uses the disk only for video recording, but there is an option to add swapspace on that disk.
This does not work well because the disk is almost always asleep (when you are just watching TV) and it is a nuisance to have it startup at random moments, or when you press some non-recording button on the remote...
In Linux, the entire kernel is unswappable.
This would probably not work as well as it did in the PDP-11 days.
Then, a program was a single file that was always loaded in memory as a whole.
Nowadays, the system uses virtual memory (so the program is not loaded all in RAM, but only the parts that are actually used are loaded).
And many programs consist of a main program and some or many dynamically loaded libraries.
Would you propose that all of these would be loaded into RAM? And would the sticky bit have to be set on each of these, or would the bit on the program itself imply that all its libraries would have to remain in RAM as well?
Don't you run the risk that newbies would set so many sticky bits that there main applications can remain loaded, but the remaining RAM is too small to accommodate the many other processes in the system, forcing these to swap and to cause disk performance problems for the main applications as well?
Some manual management capabilities can be nice, but they require competent administrators. The wide use of computers without any competent administrations has brought the (Internet) world where it is today, in an ever worsening virus and worm disaster. I don't think relying on competent administration is a good plan.
What happens (also on Linux) is that while you are away, some background process runs that reads a lot of disk data. Things like a backup, the deamon for updating the database for locating files based on name and/or content, etc.
This background process tempts the system to believe that it needs lots of pages for disk cache, and it better swaps out program memory.
After the process is finished, disk pages remain in the cache and the programs remain on swap, until you touch the mouse.
The system should examine the disk cache, the amount of free memory, the amount and age of swapped pages, and the activity of the disk.
When the disk is idle and there is a lot of free memory and/or there are a lot of cached disk sectors that are not used recently, it could copy back pages from swap to RAM (maybe leaving them valid in swap as well until they are touched in RAM).
Then, it can prepare for the user to start using the foreground programs again before he touches the mouse.
Well, in cases like this (the ARP table or some other network-related table of pending operations using up all available memory) it is just as valid to claim that the OS should limit the amount of memory spent on such things.
This depends on whether the program can do a better job at memory allocation and swapping than the operating system.
This may be true is specific situations, e.g. when the program knows it is done with operating on a specific memory page it can write it back and free the memory, whereas the OS will have to learn this by overseeing the situation.
When the addressing space allows (64-bit system) it is possible to build an OS where the entire disk is in fact a virtual address space, of which the RAM is only a "cache". Programs can just access the disk as if it were memory, and the OS will handle all the reading and writing.
With the possibility of "hints" from the application, this would probably work quite well for applications operating on lots of data.
This is not completely true. When you add some swap to your systems you will find that there is some margin where it will use the swap but not slowdown. This is because some processes allocate memory that they later (almost) never touch. This will be swapped out, and your RAM will be used for better things.
I don't know how swapping to a file performs these days, but "back in the old days" (when swap was really required because you could not afford the RAM necessary) it was noticably slower than a swap partition.
Just like you make sure that you have enough RAM, I make sure that I allocate a swap partition (on the first cylinders of the HDD) of 1-2 times RAM size. When it is not required, so be it. But my 1GB RAM system has 100MB of swapspace in use, and I have not run any very-large applications in the 41 days it is up.
Apparently it is possible to program a crappy UI on any OS, including Linux. It is also possible to build hardware that can only recover from crashes via a powercycle.
That is hardly news. It is also hardly news that it is possible to build such boxes and use Linux to good effect. TiVO and the Dreambox are examples.
It is a nice idea but it shout *not queue mail*.
The reason why individual customers may want to run their own server is to know what happens with their mail.
The mail proxy should only monitor the traffic while it relays the packets between the customer and the server he connects. When the outgoing mail stream shows virus signatures, the customer should be blocked from outgoing port 25 traffic and informed about this.
>Given the gigantic expansion of broadband, I'm surprised that cable / dsl modems don't just do NAT and other firewalling techniques by default.
They don't?? Most modems used here do!
E.g. The Alcatel SpeedTouch 510, the Draytek Vigor series. They all have built-in NAT and firewalling.
In its default configuration, it helps against Blaster and Sasser, but not against mailviruses and the resulting spam trojans.
So outgoing port 25 blocking would need to be configured by the user. The user is too clueless to do that. (if not, he would have installed a virus scanner)
>The only way to selectively allow individual ip addresses to be able to use outbond would be to have individual allow statements for each customer who requested it placed on the ACL.
Of course that is not the *only* way.
The first alternative you can think about is to have separate IP ranges of filtered and unfiltered customers, and then put the customer in the correct range depending on his blocking status.
Broadband customers normally have no guaranteed fixed IP.
They can simply block all outgoing port 25 traffic from their customers without effect on most of them.
I don't know how Comcast operates their broadband service, but here the providers often have some "service centre" page accessible to the customer, where they can configure things like pop mailbox names, virus scanning, spam filtering, etc. On such a page there could be a selection for port filtering. By default some ports will be blocked for everyone (outgoing 25, incoming 25, 80, 137-139, etc) and the customer can enable them when he thinks that is necessary.
By having the filtering on by default, all the ordinary clueless users who just want to browse and mail via the ISP mailservers will have protection from many attacks. Those who click on the enables probably know how to install a scanner and firewall.
Comcast itself could transparently proxy all web access to a server that outputs that information.
I think it is a good idea.
Furtermore, I think that Internet providers should implement a standard method for reporting infected PC's by IP address and timestamp. They can forward this message to their customer.
It is moderated as Funny, but I think he has a point.
Remember how Linux won the battle against *BSD.
It just had a better mascot and a nicer name...
Of course you will still need the phone lines!
Unless they are replaced by fiber to the home, or so. But that costs so much that rarely anyone is doing it.
Broadband is also transported over phone lines.
>In Norway the Storting (Parliament) has decided to spend $0.5 billion to build an OTA infrastructure to transmit digital TV signals. Once the new network is in place, the old analog network will be switched off. And you thought the US digital TV mandate was bad...
>My prediction is that before the new digital TV network is in place it will already by severly outdated, completely overrun by the Internet revolution.
I don't think it will be that bad, unless they take a decade to build it.
Digital TV requires broadcast bandwidth of about 300 Mbit/s for a typical channel package, or multicast bandwidth of 10-20 Mbit/s to each and every home. That is something completely different from Voice over IP!!
Such bandwidth is not in place in todays typical Internet access networks, and not everyone has or wants to have such wideband Internet access.
Even now, most providers have no multicast or broadcast capability AT ALL. Before things like TV broadcasting over Internet are even to be considered, the lowlevel network must first be made multicast-capable, so that not every viewer has to setup a private connection to some streaming server.
Worse, the way many ADSL access networks are built, with ATM throughout the network and PPPoA on top of that, even precludes the efficient use of multicast...
DVB-T is a very efficient broadcast method in such an environment, something that will be hard to beat in a wired network with point-to-point characteristics.
>f you invent a method of a frequency multiplier because you need to use one in 4Mhz steps, it is reasonable that your patent would also cover someone using an identical implementation except with 33Mhz steps
That does not sound unreasonable. But when that is the case, *why* does the original patent so explicitly specify the frequency to be 4 MHz??
I was amazed when I read that. I would not expect such an implementation detail to be specified in a patent.
It is an interesting invention to have a PLL-driven frequency multiplier as a clock source for a microprocessor (although such a PLL by itself, and its use to multiply a reference frequency by some factor, is of course a pre-existing invention).
The patent does not say anything about the input and output frequency, but it does mention the 4 MHz intermediate.
To me, it seems like an integral part of the claims. Probably one that should not have been included...
It also explicitly claims that the incoming clock is divided down to 4 MHz (this exact figure is give n more than once) and then locked to the new clock.
Intel processors do not use this frequency, they use 33 MHz.
So what is the point of all this?
When Intel are violating this patent, they at most are violating some of the claims but not all of them.
Isn't that required for it to be a patent violation?
More info:
- in which country is this event going to be held?
- what infrastructure do you have available?
going to satellite link should only be a last resort.
Apparently, major distributors (like SuSE) have also been informed, and now every user can download the update.
He is in the Netherlands (Europe).
We have some different standards and situations here.
For example, telephone poles are out of the question. The distances are probably small (dense country).
Different types of connector are normally used (IEC instead of F).
And Radio Shack moved out of here a decade ago.
But in the Netherlands it is the other way around: buying guns is illegal, and downloading music or buying drugs for your own use isn't.