So, why model against TCP w/o SACK? Isn't that just beating a paper tiger? SACK allows TCP to utilize all the packets after the lost packet(s) without retransmitting them, and requires no more memory in the transmitter and receiver than coded TCP, and with proper tuning of window size with regards to round-trip time and loss rate, should work just as well, and its been around for 25 years or so.
Well, then the coded TCP researchers excluded technology that's in modern TCP that makes their work redundant. In colloquial terms, they slayed a paper tiger. Is SACK used in current smartphones, or does a 2% or 5% erasure rate (the rates indicated in the benchmark figures of the paper) somehow prevent it from working? Omitting SACK from a cellphone's TCP implementation (or implementing it poorly) would be a cause of low transfer rates and wasted traffic in the airwaves.
It looks like RFC 1072 (TCP extensions for long-delay paths, 1996) and RFC 2018 (TCP selective Acknowlegement options, 1988) already does what I suggest. Does anyone know why it's not already widely adopted? If these RFCs were in place, the additional speed-up of coded TCP ought to be negligible.
Here's a quote from the coded TCP paper: "Other variants include those with selective acknowledgment schemes [21]. It may be interesting to compare the performance of the TCP variants with that of TCP/NC. However, we focus on traditional TCP here." The footnote [21] is to a paper that appears to be based on RFC 2018 with matching authors. I'd suggest that if they compared TCP/NC to RFC 2018, they'd see that their mechanism is way more complicated and has negligible advantages.
This mechanism has been in place for years with the PAR encoding used in USENET messages. A little redundancy permits one or more messages to be dropped without losing the transmission.
A much simpler mechanism could get most of the claimed benefit:
In TCP, a packet erasure is assumed to have resulted from congestion. However, if we start receiving packets out of order when the window size is large, we could instead presume that a packet erasure may have occurred. In a sequence of packets sent when the window size is 20 packets, say packets 1-20, packet 5 is missing, instead of NAKing packet 5 and dropping 6-20, ACKing packets 6-20+ could provoke a response of retransmitting packet 5 and once 5 is ACKed, the transmitter can move on past packet 21 (by the time packet 5 is presumed lost, the transmitter would have transmitted even more packets) - thus only lost packets are retransmitted.
The size of the transmit window can be reduced modestly when this occurs, instead of resetting it to one - perhaps the appropriate adjustment is to where the transmit window would be after successfully ACKing 19 packets after a reset. This can be generalized for multiple lost packets within the transmit window, and no "algebra" on packets would be required, only a similar amount of memory at the receiver and transmitter to the coded TCP scheme. Coded TCP takes additional bandwidth because (1) more redundant packets are sent than are needed (2) the overhead of sending which packets are algebraically combined in each packet.
The proposal assumes a constant erasure rate with packet size - for many networks this is not true. Reducing the packet size often improves the erasure rate. I've seen this in practice, in a DSL line where reducing the MTU from 1500 to a few hundred made the line usable (until I could get it fixed).
The coded TCP scheme (or the above scheme) could be implemented in software using UDP packets. Why it needs new (and licensed!?!) hardware is beyond my understanding.
Adjust the sliders to match the production increases over the last ten years, and you get 38 years left for oil, 42 years left for natural gas, and 44 years left for coal. Which makes the premise that "The World has Huge Natural Gas Reserves" totally false, unless you have no children and only expect to live for 40 years or less.
How many years of Sunlight Reserves do we have left?
Over 4,000,000,000 years.
Do you need a visualization to understand the difference between 40 years and 4,000,000,000 years?
One could spoof Carwings as a Nissan Leaf doing laps around all police stations at Mach 0.99. Sweet. And that'll play hell with the platinum ratings, especially as the battery condition will be charging instead of discharging.
http://articles.sfgate.com/1999-09-16/news/17698950_1_insurance-salesman-intelligence-test-police-chief-fred-lau
Robert Jordan -- the would-be policeman from Connecticut who scored too high on an intelligence test -- is not interested in becoming a San Francisco cop, despite a personal invitation from Police Chief Fred Lau.
"I don't think I could afford to make the move. It's not that I don't want to," Jordan said yesterday from his home in Waterford, Conn....
Lau said he was not guaranteeing a job to Jordan -- or anyone else who applies with similar intellectual vigor. But Lau said that the San Francisco Police Department, which is set to hire 200 officers in the next few months, would do well with people such as Jordan, a 48-year-old former insurance salesman.
Out of more than 2,000 officers who work for the San Francisco police, at least 10 have law degrees and a handful of others have doctorates, Lau said.
I'd rather see California do a trial of $200 netbooks running Ubuntu and accessing KahnAcademy.org and a variety of free textbooks that students could use and own for the rest of their professional life. But I guess Apple, Microsoft, and Houghton-Mifflin wouldn't profit from that. I've personally seen hundreds of bucks go down the drain buying multiple overlapping editions of Stewart Calculus for my kids.
These HFT trades are communicating something. We don't know what. It could be collusion among HFT traders. It could be communicating insider trading information. There's enough information in the signals that it could be VO-HFT, which makes sense, knowing that traders have their phone, email and IM communication recorded to assure the SEC that there's nothing illegal going on. The SEC needs to make this HFT sideband stop ASAP.
LSMFT.
The original bug in IEFBR14 was that it didn't set the exit code to zero. Fixing that bug doubled the size of the program (from one instruction to two).
Because the Kindle is not accessible to the blind because: (1) the menu system does not speak and (2) the text-to-speech system is itself disabled for some books, even when the Kindle is sold to a person with a documented, certified, reading disability. For me, as the parent of two students with reading disabilities, Amazon's Kindle will never be acceptable until they fix the second problem. Audio books from RFBD and Bookshare, have their own problems, but are much more suitable than the Kindle for the blind and dyslexic.
Rather than complaining that HTC is "dragging their feet on GPL source release," one could instead be asserting that HTC is using "unlicensed, pirated software with possible security problems."....Use the source, Luke.
....or gear the punishment to the amount of the performance claim. An eight-digit fine (that's two digits plus the other six digits) would make them take notice more than the five-digit fine actually levied.
In fact, M$ might end up settling these Vista-Capable lawsuits by offering upgrades to W7, especially if it's faster on the barely-capable hardware that the subject of the suit. Cheap way to settle for them...
It's worse than the MTBF implies. It used to be, drives would usually start failing with a few bad blocks here and there. The last rounds of failures I dealt with, the usual failure mode was loss of the entire set of data on the drive, probably due to head or controller failure. It's also made worse by the way controllers are no longer interchangable, and by placing controller firmware on the platter instead of solid-state storage. With a few bad blocks type of failure, you could always rebuild a RAID5 so long as the bad blocks on two drives don't occur at coincident locations. When the whole drive goes south at once, any failure in the other drives means data loss.
What I'd like to see specified in addition to the MTBF is the MTBF weighted by the amount of data loss. 10^-14 multiplied by drive capacity of 10^12 means a reliability index of only 10^-2, while 10^-14 multiplied by a kB of bad blocks multiplies out to 10^-11. The type of failure makes for an enormous difference.
Actually HP had an IR LED network running at HP Labs in Palo Alto around 1980 Access points were mounted in the ceiling and client nodes were mounted on poles to get over the cubicle walls.
If you look at the satellite view, there are what appear to be two enormous trees laid down on the ground. That, plus the difference between the assessor's photograph and the street-view images (which are now legally and irretrievably published in their own complaint - check out http://www.thesmokinggun.com/archive/years/2008/0407081google1.html ) suggest to me that they're pissed off that Google has exposed their tree-killing and redevelopment of the "class D" property.
Go read Sequoia's explanation. Under their scenario, voters were presented with THE WRONG CHOICES for PRESIDENT of the UNITED STATES. Specifically, Democratic party members were presented with Republican party choices or vice versa - and (this is important) - the error was not detected by the voter. This should be reason enough to disqualify the equipment!
Slashdot Mirror
So, why model against TCP w/o SACK? Isn't that just beating a paper tiger? SACK allows TCP to utilize all the packets after the lost packet(s) without retransmitting them, and requires no more memory in the transmitter and receiver than coded TCP, and with proper tuning of window size with regards to round-trip time and loss rate, should work just as well, and its been around for 25 years or so.
Well, then the coded TCP researchers excluded technology that's in modern TCP that makes their work redundant. In colloquial terms, they slayed a paper tiger. Is SACK used in current smartphones, or does a 2% or 5% erasure rate (the rates indicated in the benchmark figures of the paper) somehow prevent it from working? Omitting SACK from a cellphone's TCP implementation (or implementing it poorly) would be a cause of low transfer rates and wasted traffic in the airwaves.
Yes, it would.
It looks like RFC 1072 (TCP extensions for long-delay paths, 1996) and RFC 2018 (TCP selective Acknowlegement options, 1988) already does what I suggest. Does anyone know why it's not already widely adopted? If these RFCs were in place, the additional speed-up of coded TCP ought to be negligible.
Here's a quote from the coded TCP paper: "Other variants include those with selective acknowledgment schemes [21]. It may be interesting to compare the performance of the TCP variants with that of TCP/NC. However, we focus on traditional TCP here." The footnote [21] is to a paper that appears to be based on RFC 2018 with matching authors. I'd suggest that if they compared TCP/NC to RFC 2018, they'd see that their mechanism is way more complicated and has negligible advantages.
This mechanism has been in place for years with the PAR encoding used in USENET messages. A little redundancy permits one or more messages to be dropped without losing the transmission.
A much simpler mechanism could get most of the claimed benefit:
In TCP, a packet erasure is assumed to have resulted from congestion. However, if we start receiving packets out of order when the window size is large, we could instead presume that a packet erasure may have occurred. In a sequence of packets sent when the window size is 20 packets, say packets 1-20, packet 5 is missing, instead of NAKing packet 5 and dropping 6-20, ACKing packets 6-20+ could provoke a response of retransmitting packet 5 and once 5 is ACKed, the transmitter can move on past packet 21 (by the time packet 5 is presumed lost, the transmitter would have transmitted even more packets) - thus only lost packets are retransmitted.
The size of the transmit window can be reduced modestly when this occurs, instead of resetting it to one - perhaps the appropriate adjustment is to where the transmit window would be after successfully ACKing 19 packets after a reset. This can be generalized for multiple lost packets within the transmit window, and no "algebra" on packets would be required, only a similar amount of memory at the receiver and transmitter to the coded TCP scheme. Coded TCP takes additional bandwidth because (1) more redundant packets are sent than are needed (2) the overhead of sending which packets are algebraically combined in each packet.
The proposal assumes a constant erasure rate with packet size - for many networks this is not true. Reducing the packet size often improves the erasure rate. I've seen this in practice, in a DSL line where reducing the MTU from 1500 to a few hundred made the line usable (until I could get it fixed).
The coded TCP scheme (or the above scheme) could be implemented in software using UDP packets. Why it needs new (and licensed!?!) hardware is beyond my understanding.
Adjust the sliders to match the production increases over the last ten years, and you get 38 years left for oil, 42 years left for natural gas, and 44 years left for coal. Which makes the premise that "The World has Huge Natural Gas Reserves" totally false, unless you have no children and only expect to live for 40 years or less.
How many years of Sunlight Reserves do we have left?
Over 4,000,000,000 years.
Do you need a visualization to understand the difference between 40 years and 4,000,000,000 years?
One could spoof Carwings as a Nissan Leaf doing laps around all police stations at Mach 0.99. Sweet. And that'll play hell with the platinum ratings, especially as the battery condition will be charging instead of discharging.
http://articles.sfgate.com/1999-09-16/news/17698950_1_insurance-salesman-intelligence-test-police-chief-fred-lau Robert Jordan -- the would-be policeman from Connecticut who scored too high on an intelligence test -- is not interested in becoming a San Francisco cop, despite a personal invitation from Police Chief Fred Lau. "I don't think I could afford to make the move. It's not that I don't want to," Jordan said yesterday from his home in Waterford, Conn. ...
Lau said he was not guaranteeing a job to Jordan -- or anyone else who applies with similar intellectual vigor. But Lau said that the San Francisco Police Department, which is set to hire 200 officers in the next few months, would do well with people such as Jordan, a 48-year-old former insurance salesman.
Out of more than 2,000 officers who work for the San Francisco police, at least 10 have law degrees and a handful of others have doctorates, Lau said.
I'd rather see California do a trial of $200 netbooks running Ubuntu and accessing KahnAcademy.org and a variety of free textbooks that students could use and own for the rest of their professional life. But I guess Apple, Microsoft, and Houghton-Mifflin wouldn't profit from that. I've personally seen hundreds of bucks go down the drain buying multiple overlapping editions of Stewart Calculus for my kids.
These HFT trades are communicating something. We don't know what. It could be collusion among HFT traders. It could be communicating insider trading information. There's enough information in the signals that it could be VO-HFT, which makes sense, knowing that traders have their phone, email and IM communication recorded to assure the SEC that there's nothing illegal going on. The SEC needs to make this HFT sideband stop ASAP. LSMFT.
42. [99] Complete The_Art_of_Computer_Programming.
Who's on first?
Not many Caddies with 12,935,000,000 miles on the odometer. Plus traveling at 55,730 MpH really pegs the speedometer.
Gray hasn't got the anatomy. The Loser has the liver of a liver and a lover of the lager, not the liver of a lifer.
The original bug in IEFBR14 was that it didn't set the exit code to zero. Fixing that bug doubled the size of the program (from one instruction to two).
Because the Kindle is not accessible to the blind because: (1) the menu system does not speak and (2) the text-to-speech system is itself disabled for some books, even when the Kindle is sold to a person with a documented, certified, reading disability. For me, as the parent of two students with reading disabilities, Amazon's Kindle will never be acceptable until they fix the second problem. Audio books from RFBD and Bookshare, have their own problems, but are much more suitable than the Kindle for the blind and dyslexic.
Rather than complaining that HTC is "dragging their feet on GPL source release," one could instead be asserting that HTC is using "unlicensed, pirated software with possible security problems." ....Use the source, Luke.
Having two six-inch metal screws in my tibia has never impaired my ability to pass through an airport metal detector.
....or gear the punishment to the amount of the performance claim. An eight-digit fine (that's two digits plus the other six digits) would make them take notice more than the five-digit fine actually levied.
In fact, M$ might end up settling these Vista-Capable lawsuits by offering upgrades to W7, especially if it's faster on the barely-capable hardware that the subject of the suit. Cheap way to settle for them...
It's worse than the MTBF implies. It used to be, drives would usually start failing with a few bad blocks here and there. The last rounds of failures I dealt with, the usual failure mode was loss of the entire set of data on the drive, probably due to head or controller failure. It's also made worse by the way controllers are no longer interchangable, and by placing controller firmware on the platter instead of solid-state storage. With a few bad blocks type of failure, you could always rebuild a RAID5 so long as the bad blocks on two drives don't occur at coincident locations. When the whole drive goes south at once, any failure in the other drives means data loss. What I'd like to see specified in addition to the MTBF is the MTBF weighted by the amount of data loss. 10^-14 multiplied by drive capacity of 10^12 means a reliability index of only 10^-2, while 10^-14 multiplied by a kB of bad blocks multiplies out to 10^-11. The type of failure makes for an enormous difference.
Actually HP had an IR LED network running at HP Labs in Palo Alto around 1980 Access points were mounted in the ceiling and client nodes were mounted on poles to get over the cubicle walls.
...to clarify, this comment is regarding the Boring's, not the McKee's.
If you look at the satellite view, there are what appear to be two enormous trees laid down on the ground. That, plus the difference between the assessor's photograph and the street-view images (which are now legally and irretrievably published in their own complaint - check out http://www.thesmokinggun.com/archive/years/2008/0407081google1.html ) suggest to me that they're pissed off that Google has exposed their tree-killing and redevelopment of the "class D" property.
Go read Sequoia's explanation. Under their scenario, voters were presented with THE WRONG CHOICES for PRESIDENT of the UNITED STATES. Specifically, Democratic party members were presented with Republican party choices or vice versa - and (this is important) - the error was not detected by the voter. This should be reason enough to disqualify the equipment!
I also note the use of the term "Democrat party" in Sequoia's explanation - I'll reserve my own comments on this slur, instead referring the reader to http://en.wikipedia.org/wiki/Democrat_Party_(phrase)