There is have focus on throughput in these benchmarks. Reading and writing lots of data, seeking in files and reading data, etc.
Notably missing are more day-to-day useful operations such as the creation and deletion of lots of files, parallel action on many open files, lots of files in a directory, etc.
When I want to select a filesystem, I do not want to know how fast it can read a 3GB file sequentially. I want to know how well it performs on a fileserver, mailserver etc.
Yes, this is a very weak link in that system. I have seen it on other mini systems as well. When the console is off or defective, or Ctrl-S or scroll lock is hit, the whole system stops (after a while) or won't boot. We have had our AIX box hanging during a weekly nightly reboot because someone switched off the console terminal. Indeed it can leave you puzzled for a while, especially as this console is rarely used, and operators normally use network connections to access the machine.
While I think that sitefinder is the wrong way to innovate (it caused me problems as well), I do think that DNS needs some innovation.
The hierarchical model of DNS names was invented by techies, for reasons known and understood by techies. Today, the Internet is used by people on the street, and they do not understand (or want to spend effort to understand) this hierarchical model. They just see a name like "company.com" as a reference to that company, and do not know why the.com part is there and it sometimes is.org or.net or.countrycode or whatever.
The first apparent failure of the hierarchical model has been that everyone registers productnames, advertisement slogans, and whatever you can imagine as a second-level domain under.com or a country code. This has never been the intention. Registrations on the second level should have been limited.
Now that extra top-level domains appear, this failure is only multiplied. Every business that already registered a name now also registers (or should register) a.biz domain with the same name. Or with the same 100 names they already held.
It is very clear that this hierarchical structure does not work in the real world. DNS needs an innovation to remove this dotted-name system and replace it by something more intuitive, and less prone to abusive registration of many names referring to the same thing.
Finding a form that this should take is of course quite a challenge. It goes beyond a silly innovation like registering a wildcard domain.
You for get that most modern boards have several separate USB controllers. Who would ever want to connect keyboard, mouse, joystick, mp3 player, external drive off a single controller using a HUB???
The minumom number of controllers I have seen on today's machines is three (two UHCI and one EHCI).
Yes, the different keys produce different codes and the meaning is defined by the drivers (or in DOS: the BIOS when no specific KEYB driver is loaded).
The original keyboard had no right Alt key. Maybe some early clone keyboards with two Alt keys had them wired in parallel. But the later 101-key keyboards with two Alt keys always had this AltGr key on the international versions, and indeed in some drivers they will not do the same thing and Ctrl-AltGr-Del does nothing.
Here in the Netherlands these country-specific keyboards have almost ceased to exist, and US keyboards are now sold by default for nearly all computers. However, they do have a EURO symbol that can be accessed via the right Alt key and the digit 5...
This kind of services often does not make it because of the payment problem. You want the sender to pay for the handling (just as with regular mail), because if the receiver pays they will have no protection against spam, abuse, etc. But there is no sender-pays infrastructure in place in the e-mail system. Once that has been built, it would not only be possible to implement this kind of service, but it would also be the solution for the spam problem. Of course, the sender would get some information about the tariff when sending a message. A normal e-mail message may cost 1 cent or so, and a post-delivered message could be 50 cents. When you agree to that at the moment of sending the message, that would be perfectly OK. (spammers of course would not agree, and would no longer be able to send you mail)
The marking on the key has nothing to do with the functioning. All PC keyboards send the same codes, independent of the printing on the keycaps. If the right Alt key is different from the left is dependent on the keyboard layout (country) selected, and of course also on the operating system.
Not quite... The keyboard has a serial interface to the system. On the AT, this serial interface is handled by a small microcontroller. This microcontroller accepts instructions from the 286 via two I/O ports. There is a special instruction that tells the microcontroller to pull the RESET line on the 286. This is used (together with a special value in the RTC NVRAM) to signal a return from protected mode.
After this became really used, the slowness of the microcontroller quicly became a hindrance, and special hardware was added in clones and later models to quickly decode this instruction at the moment it was output to the I/O port and immediately reset the CPU instead of waiting for the microcontroller to act.
In the days the CTRL-ALT-DEL was invented, there were no Alt and Ctrl keys to the right side of the spacebar, and there was no Insert/Home/PgUp/Delete/End/PgDown keygroup. The only DEL was the one on the numeric keypad.
I think this is not how it works, at least not today. These engine systems take a tank of gas, then accellerate this gas using some ionosation device (like a spark) and get their thrust from that. The only advantage is that they can get a high exhaust speed and thus a highly efficient conversion from mass into impulse. (and part of the energy this uses is taken from the solar panels while the craft is underway, instead of being stored in the fuel all the time)
However, this is not an infinite source of thrust. Once the has tank is empty, it stops.
That is not going to work with as many ISPs as there are now. There are always going to be ISPs who are in the grey zone, claiming they are serving only legitimate customers but doing not enough to keep out the spammers. E.g. ISPs that offer free dialup accounts, or even anonymous dialup. There is no simple "kicking them out", there will have to be a time-consuming procedure of warning, warning again, and finally maybe a disconnection that will be subject to appeal etc.
A "secure network" could be practical with 10 or 100 parties, but not with 100.000 or 1.000.000
Last time I built a new system with two 120GB disks I thought "well let's consider to use LVM", after having seen it on IBM AIX.
So, I started experimenting with SuSE Linux installation on LVM. Of course I wanted RAID-1. Well, that did not seem to be considered. You could make two partitions, RAID-1 them and run LVM on top of that, but no RAID-1 as part of the LVM (as in AIX).
Next, it turned out that you cannot have root on LVM. And no swap either.
So, I would need to make 3 partitions on both drives, make 3 RAID-1 sets, use one for root one for swap and the remainder could then finally be used for LVM. At that time, I decided that it was not worth it (for my home machine)
When you have LVM, it should be a total solution, not something you can only use for/var and/home.
(after some time I found out that my root partition is too large and the remainder-of-the-disk partition that I mounted as/local and have put things like/home and/usr/local in is too small. LVM would have helped if it was good, but it would not have helped me because of the restrictions I mentioned)
Conclusion: LVM is a good idea but the implementation in Linux needs more work.
Are you sure this same model isn't used in "the other camp"?
I mean, what do I buy for having "Designed for Windows XP" stickers on my system that of course runs Linux? It must cost money to do the necessary validations and to attach the stickers during manufacture.
At the end of the day, the customer is paying for that, even when he does not want to run Windows. (worse, he may be even paying for a Windows license he is not going to use)
Re:Great for GC and dynamically typed languages
on
Athlon 64 Debuts
·
· Score: 1
These managed-run-time/GC/DT languages are pointer heavy. Your pointers just doubled in size with 64 bits. Therefore, your caches can't hold as much data as before and also your memory bandwidth is narrowed because you need to transfer more data across the bus.
That is a general problem with 64-bit processors. When nothing else changes, they tend to be slower.
Re:Great for GC and dynamically typed languages
on
Athlon 64 Debuts
·
· Score: 1
What you really want is a tag field that is outside of the value bits of the registers and memory, like the Burroughs B7000 mainframes had. This controlled whether a word contained an instruction, an integer, a float, a pointer, etc. The instruction size was quite small on these processors as data type information was retrieved from the value, not the instruction. It also provided protection against accidental execution of data or overwrite of stack frames.
Unfortunately their first cut at it (P1) was buggy. I have run it for a few days and it causes errors on perfectly valid lookups (and corresponding entries in the logfile).
Now there seems to be a P2 that is supposed to fix that, but it is only available as a full download, not a patch to P1 or the bare version. Impractical for me, as I already have the sourcetree with other patches (from distributor) on the system and only want to add the fix.
So the only option for now was to remove P1 and go back to the standard version. I wonder how many ISPs have installed this and either stick at P1 (with bugs) or have to do extra work to make it correct.
When you pay your taxes to use the road, or when you buy a season-ticket to a bus or train line, you by the rights to use this road, bus or train as much as you like. And when you ride back and forth all day, nobody will care or even notice.
But when *everybody* would do that, capacity of the road, bus or train would be insufficient. That is because such capacities are calculated based on average usage patterns. Experience shows that an average train passenger does not ride back and forth all day, but only once or twice.
Experience also shows that a broadband user users maximum capacity for only 2-4% of the time, or uses only 2-4% of maximum capacity all the time. So, it is possible to have 25-40 users of the same bandwidth without anybody noticing, unless something special happens and they all want to use the capacity at the same time.
When you don't want to be subjected to such statistical capacity calculations, you can say so when you get your connection, and you will be put into another user category (business user with specified bandwidth guarantee).
Slashdot Mirror
Well, another big player in the Linux field, SuSE, uses Reiserfs by default.
There is have focus on throughput in these benchmarks. Reading and writing lots of data, seeking in files and reading data, etc.
Notably missing are more day-to-day useful operations such as the creation and deletion of lots of files, parallel action on many open files,
lots of files in a directory, etc.
When I want to select a filesystem, I do not want to know how fast it can read a 3GB file sequentially. I want to know how well it performs on a fileserver, mailserver etc.
That was even worse! It introduced another failure mode: out of paper on the console.
I'll bet your turbo key alternates between sending Control-Alt-NumberPadPlus and Control-Alt-NumberPadMinus.
:-)
That would be nice on a XFree86 system
Yes, this is a very weak link in that system. I have seen it on other mini systems as well. When the console is off or defective, or Ctrl-S or scroll lock is hit, the whole system stops (after a while) or won't boot.
We have had our AIX box hanging during a weekly nightly reboot because someone switched off the console terminal.
Indeed it can leave you puzzled for a while, especially as this console is rarely used, and operators normally use network connections to access the machine.
While I think that sitefinder is the wrong way to innovate (it caused me problems as well), I do think that DNS needs some innovation.
.com part is there and it sometimes is .org or .net or .countrycode or whatever.
.com or a country code. This has never been the intention. Registrations on the second level should have been limited.
.biz domain with the same name. Or with the same 100 names they already held.
The hierarchical model of DNS names was invented by techies, for reasons known and understood by techies. Today, the Internet is used by people on the street, and they do not understand (or want to spend effort to understand) this hierarchical model.
They just see a name like "company.com" as a reference to that company, and do not know why the
The first apparent failure of the hierarchical model has been that everyone registers productnames, advertisement slogans, and whatever you can imagine as a second-level domain under
Now that extra top-level domains appear, this failure is only multiplied. Every business that already registered a name now also registers (or should register) a
It is very clear that this hierarchical structure does not work in the real world.
DNS needs an innovation to remove this dotted-name system and replace it by something more intuitive, and less prone to abusive registration of many names referring to the same thing.
Finding a form that this should take is of course quite a challenge. It goes beyond a silly innovation like registering a wildcard domain.
You for get that most modern boards have several separate USB controllers. Who would ever want to connect keyboard, mouse, joystick, mp3 player, external drive off a single controller using a HUB???
The minumom number of controllers I have seen on today's machines is three (two UHCI and one EHCI).
You should have read the claims of the patent and you would have found that it does not describe a system like that.
Yes, the different keys produce different codes and the meaning is defined by the drivers (or in DOS: the BIOS when no specific KEYB driver is loaded).
The original keyboard had no right Alt key. Maybe some early clone keyboards with two Alt keys had them wired in parallel. But the later 101-key keyboards with two Alt keys always had this AltGr key on the international versions, and indeed in some drivers they will not do the same thing and Ctrl-AltGr-Del does nothing.
Here in the Netherlands these country-specific keyboards have almost ceased to exist, and US keyboards are now sold by default for nearly all computers. However, they do have a EURO symbol that can be accessed via the right Alt key and the digit 5...
This kind of services often does not make it because of the payment problem. You want the sender to pay for the handling (just as with regular mail), because if the receiver pays they will have no protection against spam, abuse, etc.
But there is no sender-pays infrastructure in place in the e-mail system.
Once that has been built, it would not only be possible to implement this kind of service, but it would also be the solution for the spam problem.
Of course, the sender would get some information about the tariff when sending a message. A normal e-mail message may cost 1 cent or so, and a post-delivered message could be 50 cents. When you agree to that at the moment of sending the message, that would be perfectly OK.
(spammers of course would not agree, and would no longer be able to send you mail)
Non-native english writers often forget that "for example" is abbreviated as "e.g.", not "f.e." or "for ex." or whatever...
The marking on the key has nothing to do with the functioning. All PC keyboards send the same codes, independent of the printing on the keycaps.
If the right Alt key is different from the left is dependent on the keyboard layout (country) selected, and of course also on the operating system.
Not quite...
The keyboard has a serial interface to the system. On the AT, this serial interface is handled by a small microcontroller. This microcontroller accepts instructions from the 286 via two I/O ports.
There is a special instruction that tells the microcontroller to pull the RESET line on the 286. This is used (together with a special value in the RTC NVRAM) to signal a return from protected mode.
After this became really used, the slowness of the microcontroller quicly became a hindrance, and special hardware was added in clones and later models to quickly decode this instruction at the moment it was output to the I/O port and immediately reset the CPU instead of waiting for the microcontroller to act.
In the days the CTRL-ALT-DEL was invented, there were no Alt and Ctrl keys to the right side of the spacebar, and there was no Insert/Home/PgUp/Delete/End/PgDown keygroup. The only DEL was the one on the numeric keypad.
JKL = screen print
DFG = DOS command prompt while running another program
123 = debugger
I think this is not how it works, at least not today.
These engine systems take a tank of gas, then accellerate this gas using some ionosation device (like a spark) and get their thrust from that.
The only advantage is that they can get a high exhaust speed and thus a highly efficient conversion from mass into impulse.
(and part of the energy this uses is taken from the solar panels while the craft is underway, instead of being stored in the fuel all the time)
However, this is not an infinite source of thrust. Once the has tank is empty, it stops.
That is not going to work with as many ISPs as there are now.
There are always going to be ISPs who are in the grey zone, claiming they are serving only legitimate customers but doing not enough to keep out the spammers.
E.g. ISPs that offer free dialup accounts, or even anonymous dialup.
There is no simple "kicking them out", there will have to be a time-consuming procedure of warning, warning again, and finally maybe a disconnection that will be subject to appeal etc.
A "secure network" could be practical with 10 or 100 parties, but not with 100.000 or 1.000.000
And how are you going to certify the ISPs allowed on that network, so you won't get any spammers on the list of acceptable senders?
It looked good at first but the Yast2 installer (SuSE) disallowed both root on LVM and swap on LVM...
Hopefully it will be improved in the future.
Last time I built a new system with two 120GB disks I thought "well let's consider to use LVM", after having seen it on IBM AIX.
/var and /home.
/local and have put things like /home and /usr/local in is too small. LVM would have helped if it was good, but it would not have helped me because of the restrictions I mentioned)
So, I started experimenting with SuSE Linux installation on LVM. Of course I wanted RAID-1. Well, that did not seem to be considered. You could make two partitions, RAID-1 them and run LVM on top of that, but no RAID-1 as part of the LVM (as in AIX).
Next, it turned out that you cannot have root on LVM. And no swap either.
So, I would need to make 3 partitions on both drives, make 3 RAID-1 sets, use one for root one for swap and the remainder could then finally be used for LVM. At that time, I decided that it was not worth it (for my home machine)
When you have LVM, it should be a total solution, not something you can only use for
(after some time I found out that my root partition is too large and the remainder-of-the-disk partition that I mounted as
Conclusion: LVM is a good idea but the implementation in Linux needs more work.
Are you sure this same model isn't used in "the other camp"?
I mean, what do I buy for having "Designed for Windows XP" stickers on my system that of course runs Linux?
It must cost money to do the necessary validations and to attach the stickers during manufacture.
At the end of the day, the customer is paying for that, even when he does not want to run Windows.
(worse, he may be even paying for a Windows license he is not going to use)
These managed-run-time/GC/DT languages are pointer heavy. Your pointers just doubled in size with 64 bits. Therefore, your caches can't hold as much data as before and also your memory bandwidth is narrowed because you need to transfer more data across the bus.
That is a general problem with 64-bit processors.
When nothing else changes, they tend to be slower.
What you really want is a tag field that is outside of the value bits of the registers and memory, like the Burroughs B7000 mainframes had.
This controlled whether a word contained an instruction, an integer, a float, a pointer, etc.
The instruction size was quite small on these processors as data type information was retrieved from the value, not the instruction.
It also provided protection against accidental execution of data or overwrite of stack frames.
Unfortunately their first cut at it (P1) was buggy. I have run it for a few days and it causes errors on perfectly valid lookups (and corresponding entries in the logfile).
Now there seems to be a P2 that is supposed to fix that, but it is only available as a full download, not a patch to P1 or the bare version. Impractical for me, as I already have the sourcetree with other patches (from distributor) on the system and only want to add the fix.
So the only option for now was to remove P1 and go back to the standard version. I wonder how many ISPs have installed this and either stick at P1 (with bugs) or have to do extra work to make it correct.
When you pay your taxes to use the road, or when you buy a season-ticket to a bus or train line, you by the rights to use this road, bus or train as much as you like. And when you ride back and forth all day, nobody will care or even notice.
But when *everybody* would do that, capacity of the road, bus or train would be insufficient. That is because such capacities are calculated based on average usage patterns. Experience shows that an average train passenger does not ride back and forth all day, but only once or twice.
Experience also shows that a broadband user users maximum capacity for only 2-4% of the time, or uses only 2-4% of maximum capacity all the time.
So, it is possible to have 25-40 users of the same bandwidth without anybody noticing, unless something special happens and they all want to use the capacity at the same time.
When you don't want to be subjected to such statistical capacity calculations, you can say so when you get your connection, and you will be put into another user category (business user with specified bandwidth guarantee).