No, US was a Brittish penal colony, just like Australia.
Australia was second. England started sending convicts over to Australia only after large enough numbers of ex-convicts (==free men) in America started to get opiniated about the "free labor" / "slavery" scheme that the penal colonies actually were and that this "free labor" threatened their own employability.
Again, once Australia had large enough numbers of "free men" as well that also became opiniated about the "free labor" threatening their jobs that the penal system consisted of then they started shipping them to Tassie instead.
US, Australia and Tassie were all originally Brittish penal colonies. Australia once they could not ship convicts to the USD any more. Tassie once they couldnt send them to Australia any more.
You can very easily transmit enough power to power a PC over air today. No problem.
The only issue is that you would likely not ever be able to have kids, though that would not be an issue for/. readers. Your expected lifespan might also be not-insignificately reduced, but it is possible.
Peter Jacksons first movie was BadTaste. DeadAlive, as it was called in the US, or Braindead as the rest of us know it was his second movie.
BadTaste was much better than DeadAlive and is a tragedy for PeterJackson. When the first movie you ever made was an 11, whewre can you go from there? You cant go for one splattier, 11 is the max.
Evil Dead might be a classic but BadTaste is the splattermovie to end all splattermovies.
If the sorting algorithm is O(n log n) then it should (for many kinds of data) always fail compared to a normal CPU running a proper sorting algorithm. IF the sort keys allow a non-compare sort algorithm.
Devicing an O(n) algorithm that does not do an expensive branch to a compare function is pretty trivial (e.g. radix sort) and will always win by a huge margin to general purpose algorithms such as qsort.
qsort is a good O(n log n) algorithm that can sort any kind of data. The only drawback with O(n) algorithms is that they can only sort sertain kinds of data, but they are much faster.
Adaptec + TOE ? it wasnt the 7212 iSCSI HBA was it?
question? did you guys ever fix the data corruption issue where the hba reassembled multiple data-in pdu's incorrectly with an 8 byte overlap between segments (and filling out the missing 8 tail bytes out with uninitialized data)?
also, did you fix the issue where the hba would only do 512 bytes of i/o every roundtrip, making high-latency comms "slow" ?
anyone that want to test the "slowness" can just try to set max tcp window-size to 512 bytes. enjoy.
to try the datacorruption issue, just try
dd if=/dev/random of=/dev/hda
By the way, RFC2581 was published in april 1999, this rfc even explains why this rfc sucks and why rfc2582 should be used. tell your friends not yet laid off to read rfc2582.
a tcp checksum is essentially a 16 bit add with overflow wrapped back to the LSB. there is no way ever that is going to consume any measurable amount of cpu compared to actually processing the data.
The least of the worries for these scientists is getting their funding revoked. They have much more pressing reasons to "be observant, be alert, dont walk into dark alleys at night" as far as unwanted interactions with DeBeer goes.
There is a reason why the DeBeer family can not travel to places like the EU or the US.
My kids will start their career like I did, get a proper degree in science like math or physics.
Word or other data-entry related skills are then used by those providing cheap and basic data-entry and burger-flipping services for my kids.
I do think it is very good that the western world is dumbing down the education system. Very good indeed, the more people are trained to do basic data-entry instead of real work the less competition my kids will have.
Microsoft used to use DCE/RPC over SMB or in dce-speak over nca-cn-pipe over well known endpoints down in \PIPE on the C$ or the same special files down in IPC$.
This had the benefit it piggybacked on the SMB authentication since it was accessed through authenticated named pipes.
All these endpoints were also available through raw TCP ports, but that wqould require explicit EPM calls and explicit authentication in the DCE/RPC layer, so it is just easier to map the IPC$ share (and automatically authenticate to the server) and then open the named pipe with the well-known magic name.
SMB related interfaces have always been available also through TCP portsd in addition to the well known pipes, try some of Todd's tools over at bindview they can enumerate all endpoints where a interface is registered.
SMB related services always used to have a well known pipe, these are services such as NETLOGON, SAMR, LSA, WKSSVC, SRVSVC, SVCCTL, etc.
Non cifs/smb related services do not use named pipes but always use the EPM to find the port to talk to (or talk through the service through the EPM endpoint). This includes the two dce protocols used by Exchange.
Starting with w2k3 Microsoft has stopped using named pipes also for normal cifs services and have started talking dce over tcp directly even for most of the classic interfaces such as samr and lsa.
At least when w2k3 boxens talk to eachother they seem to now prefer nca_cn_tcp instead of nca_cn_pipe, even for the interfaces with well known pipe endpoints.
i have no idea why and have never been curious enough to try to find out
Encrypted sessions is merely just applying GSS-KRB to DCE/RPC.
This is not a CIFS/SMB extension since it technically has nothing to do with CIFS/SMB. (eventhough CIFS/SMB is often a transport for carryiung DCE/RPC on older windows boxens, moderns ones transport DCE/RPC over raw TCP)
The only extension here that is "proprietary" would be that windows prefer to use ARCFOUR encryption due to the way it hashes passwords into secrets.
While ARCFOUR is technically "propriatary" since it has no RFC, it is a well known algorithm and many many people has implemented it. Even MIT and Heimdal Kerberos has supported ARCFOUR since ages, though they do not support GSS-KRB for DCE/RPC (yet).
Encrypted sessions is not that difficult, it is just that since is mainly used/enforced today between DCs so no one has really bothered to implement it yet in other implementations.
For someone that is bothered enough to implement "encrypted sessions" in a CIFS client/server it should not take very many days/weeks to implement/test, it is pretty simple (at elast if you only aim for ARCFOUR which is 99% of all encrypted DCE/RPC anyways).
It only took me a couple of days spare time to implement decryption of them in ethereal...
That kind of massive performance drop can really only be caused by any of three things things:
1, congestion in the switch and packetloss ==> performance degradation due to the sender sitting idle most of the time to wait for a TCP retransmission to trigger.
2, duplex mismatch causing packetloss ==> performance degradation due to the sender sitting idle most of the time to wait for a TCP retransmission to trigger.
3, if you get poor results only when uploading files to the mac but good performance when downloading, a recent buglet in some versions of the mac os where it would stop sending out a tcp ack for every full sized segment thus causing the tcp layer to cap throughput.
i would recommend running a sniffer on the network and find out why it is so slow.
Wrong, or our information is outdated.
There are multiple AD implementations that allow Microsoft clients to join to and authenticate to as if they were w2k3 AD.
See www.padl.com for one of the first implementations of one such AD. Samba4 also supports this but is a different implementation.
In general and assuming a non broken cache architecture,
a 2CPU/core solution will feel faster than a single cpu solution with twice the cpu frequency.
The total number of cpy cycles are the same, but the average queue-length for a process waiting for the CPU is half, i.e. the latency before your process is scheduled is lower making it "feel" faster.
If your os has a scheduler that moves a process back and forth between different cpus/cores with no understanding if the cost of invalidating and reloading the cache then you sould upgrade to an os that does not have a broken scheduler.
any scheduler that can not handle this is broken. do not use broken os or broken schedulers if you have more than 1 core/cpu.
That is just so likely. Maybe you went to school outside of europe based on your capability of logical reasoning?
So, lets pull out from a 25b market just to teach those dirty europeans a lesson.
Walking away from this market in pride is in the shareholders best interest?
You know that there are laws that say that the board have to act in the best interest of the shareholders?
Walking away from a 25b market and the inevitable plunge in share price is in the shareholders best interest?
I think it is fair to assume that no matter what, ms will NEVER walk away from the european market unless the board members want to be targeted by massive class-act lawsuits the size of the world has not seen before and spend significant time in jail.
the stupidity of/. posters never ceases to surprise you.
Slashdot Mirror
There is a lot of innovation going on in TCP even today.
...
While the packet formats has not changed very much there is an incredible amount of research and development and progress in TCP algorithms.
Read any papers by Sally floyd, google for TCP pacing, google for RED,
A lot of things have happened with TCP over the years and a lot of new things are happening continously even today.
If you think TCP is an old stale protocol you are mistaken.
No, US was a Brittish penal colony, just like Australia.
Australia was second. England started sending convicts over to Australia only after large enough numbers of ex-convicts (==free men) in America started to get opiniated about the "free labor" / "slavery" scheme that the penal colonies actually were and that this "free labor" threatened their own employability.
Again, once Australia had large enough numbers of "free men" as well that also became opiniated about the "free labor" threatening their jobs that the penal system consisted of then they started shipping them to Tassie instead.
US, Australia and Tassie were all originally Brittish penal colonies. Australia once they could not ship convicts to the USD any more. Tassie once they couldnt send them to Australia any more.
BS
/. readers. Your expected lifespan might also be not-insignificately reduced, but it is possible.
You can very easily transmit enough power to power a PC over air today. No problem.
The only issue is that you would likely not ever be able to have kids, though that would not be an issue for
Salon got it wrong.
Peter Jacksons first movie was BadTaste. DeadAlive, as it was called in the US, or Braindead as the rest of us know it was his second movie.
BadTaste was much better than DeadAlive and is a tragedy for PeterJackson.
When the first movie you ever made was an 11, whewre can you go from there? You cant go for one splattier, 11 is the max.
Evil Dead might be a classic but BadTaste is the splattermovie to end all splattermovies.
well, radix sort can not be used for general purpose sorting. that is the drawback, the benefit is that it IS fast.
As for sorting integers or floats, radix sort IS O(n).
rs also has nice properties that it is best-case/average-case/worst-case O(n).
qsort is only O(n log n) for average case IFF all the input data to sort is completely random or carefully selected to perform well.
try qsort() and compare it with non-random real-world data and see how they differ.
If the sorting algorithm is O(n log n) then it should (for many kinds of data) always fail compared to a normal CPU running a proper sorting algorithm. IF the sort keys allow a non-compare sort algorithm.
Devicing an O(n) algorithm that does not do an expensive branch to a compare function is pretty trivial (e.g. radix sort) and will always win by a huge margin to general purpose algorithms such as qsort.
qsort is a good O(n log n) algorithm that can sort any kind of data.
The only drawback with O(n) algorithms is that they can only sort sertain kinds of data, but they are much faster.
Adaptec + TOE ? it wasnt the 7212 iSCSI HBA was it?
question? did you guys ever fix the data corruption issue where the hba reassembled multiple data-in pdu's incorrectly with an 8 byte overlap between segments (and filling out the missing 8 tail bytes out with uninitialized data)?
also, did you fix the issue where the hba would only do 512 bytes of i/o every roundtrip, making high-latency comms "slow" ?
anyone that want to test the "slowness" can just try to set max tcp window-size to 512 bytes. enjoy.
to try the datacorruption issue, just try
dd if=/dev/random of=/dev/hda
By the way, RFC2581 was published in april 1999, this rfc even explains why this rfc sucks and why rfc2582 should be used. tell your friends not yet laid off to read rfc2582.
SAN does not share and but can cache data on clients.
NAS can share but can not cache on clients.
So exactly why is it surprising that NAS access requires more resources than SAN ?
be serious.
a tcp checksum is essentially a 16 bit add with overflow wrapped back to the LSB.
there is no way ever that is going to consume any measurable amount of cpu compared to actually processing the data.
If you had an incident you reformat the drives, all of them, and reinstall from scratch. Every time. No exception.
Doesnt matter what OS you run, after an incident you will reformat everything and start from scratch.
so fog is very efficient at blocking gamma radiation then?
Debeers will survive, yes probably.
The burning question though is, will these scientists?
The least of the worries for these scientists is getting their funding revoked. They have much more pressing reasons to "be observant, be alert, dont walk into dark alleys at night" as far as unwanted interactions with DeBeer goes.
There is a reason why the DeBeer family can not travel to places like the EU or the US.
nah
It was the failures at the Sellafield, Harrisburg, Chernobyl kraftwerks that scared people off.
My kids will start their career like I did, get a proper degree in science like math or physics.
Word or other data-entry related skills are then used by those providing cheap and basic data-entry and burger-flipping services for my kids.
I do think it is very good that the western world is dumbing down the education system. Very good indeed, the more people are trained to do basic data-entry instead of real work the less competition my kids will have.
I feel sorry for your kids if you have no higher goals for them than being able to get a job where Windows data-entry skills are a requirement.
"do they make SRVSVC available over raw TCP, and, if so, do newer versions of the Windows SMB client code use that to enumerate shares"
Yes, this service has always been available through TCP but has not been used by common clients).
Yes, when w2k3 boxens talk to eachothers nowadays, they do use nca_cn_tcp to enumerate shares.
Microsoft used to use DCE/RPC over SMB or in dce-speak over nca-cn-pipe over well known endpoints down in \PIPE on the C$ or the same special files down in IPC$.
This had the benefit it piggybacked on the SMB authentication since it was accessed through authenticated named pipes.
All these endpoints were also available through raw TCP ports, but that wqould require explicit EPM calls and explicit authentication in the DCE/RPC layer, so it is just easier to map the IPC$ share (and automatically authenticate to the server) and then open the named pipe with the well-known magic name.
SMB related interfaces have always been available also through TCP portsd in addition to the well known pipes, try some of Todd's tools over at bindview they can enumerate all endpoints where a interface is registered. SMB related services always used to have a well known pipe, these are services such as NETLOGON, SAMR, LSA, WKSSVC, SRVSVC, SVCCTL, etc.
Non cifs/smb related services do not use named pipes but always use the EPM to find the port to talk to (or talk through the service through the EPM endpoint). This includes the two dce protocols used by Exchange.
Starting with w2k3 Microsoft has stopped using named pipes also for normal cifs services and have started talking dce over tcp directly even for most of the classic interfaces such as samr and lsa. At least when w2k3 boxens talk to eachother they seem to now prefer nca_cn_tcp instead of nca_cn_pipe, even for the interfaces with well known pipe endpoints.
i have no idea why and have never been curious enough to try to find out
Nothing really proprietary here...
Encrypted sessions is merely just applying GSS-KRB to DCE/RPC.
This is not a CIFS/SMB extension since it technically has nothing to do with CIFS/SMB. (eventhough CIFS/SMB is often a transport for carryiung DCE/RPC on older windows boxens, moderns ones transport DCE/RPC over raw TCP)
The only extension here that is "proprietary" would be that windows prefer to use ARCFOUR encryption due to the way it hashes passwords into secrets.
While ARCFOUR is technically "propriatary" since it has no RFC, it is a well known algorithm and many many people has implemented it. Even MIT and Heimdal Kerberos has supported ARCFOUR since ages, though they do not support GSS-KRB for DCE/RPC (yet).
Encrypted sessions is not that difficult, it is just that since is mainly used/enforced today between DCs so no one has really bothered to implement it yet in other implementations.
For someone that is bothered enough to implement "encrypted sessions" in a CIFS client/server it should not take very many days/weeks to implement/test, it is pretty simple (at elast if you only aim for ARCFOUR which is 99% of all encrypted DCE/RPC anyways).
It only took me a couple of days spare time to implement decryption of them in ethereal...
That kind of massive performance drop can really only be caused by any of three things things:
1, congestion in the switch and packetloss ==> performance degradation due to the sender sitting idle most of the time to wait for a TCP retransmission to trigger.
2, duplex mismatch causing packetloss ==> performance degradation due to the sender sitting idle most of the time to wait for a TCP retransmission to trigger.
3, if you get poor results only when uploading files to the mac but good performance when downloading, a recent buglet in some versions of the mac os where it would stop sending out a tcp ack for every full sized segment thus causing the tcp layer to cap throughput.
i would recommend running a sniffer on the network and find out why it is so slow.
Wrong, or our information is outdated.
There are multiple AD implementations that allow Microsoft clients to join to and authenticate to as if they were w2k3 AD.
See www.padl.com for one of the first implementations of one such AD. Samba4 also supports this but is a different implementation.
In general and assuming a non broken cache architecture, a 2CPU/core solution will feel faster than a single cpu solution with twice the cpu frequency.
The total number of cpy cycles are the same, but the average queue-length for a process waiting for the CPU is half, i.e. the latency before your process is scheduled is lower making it "feel" faster.
If your os has a scheduler that moves a process back and forth between different cpus/cores with no understanding if the cost of invalidating and reloading the cache then you sould upgrade to an os that does not have a broken scheduler. any scheduler that can not handle this is broken. do not use broken os or broken schedulers if you have more than 1 core/cpu.
I am pretty good at reverse engineering network protocols.
I suddenly feel a strong urge to reverse engineer the BK protocol and add it to ethereal.
That is just so likely. Maybe you went to school outside of europe based on your capability of logical reasoning? So, lets pull out from a 25b market just to teach those dirty europeans a lesson. Walking away from this market in pride is in the shareholders best interest? You know that there are laws that say that the board have to act in the best interest of the shareholders? Walking away from a 25b market and the inevitable plunge in share price is in the shareholders best interest? I think it is fair to assume that no matter what, ms will NEVER walk away from the european market unless the board members want to be targeted by massive class-act lawsuits the size of the world has not seen before and spend significant time in jail. the stupidity of /. posters never ceases to surprise you.