I don't know what definition of "clock skew" he is using but to say you can measure clock skew by looking at time stamps generated from a single point makes no sense at all. Clock skew is the mean time difference between *two* clocks of the same frequency (or related by a rational multiplier). The skew of any clock measured against itself at the same point in a clock tree is always zero.
It seems likely to me that he is measuring the differences in clock frequency from a reference frequency. This is not clock skew. Call it "clock frequency variance" or something to that effect.
So why do you typically get three Wilco songs in an hour while Aretha Franklin waits in the wings forever?
Yep, mine plays "Less Than You Think" over and over.
OK, I've already made one pedantic comment today, but I'll risk another.
There is definitely not another Mersenne prime between the 41st and 42nd Mersenne primes.
I don't know if you meant to say something else, but you cannot "easily add more lanes to each slot". If the root complex or switch driving that slot has 4 lanes, 4 lanes is all you will get from that slot; if it has 8 lanes, 8 lanes is all you will get.
it's primarily a video interface on the desktop.
No.
It *is* replacing AGP, but it isn't limited to that.
PCI Express devices are being developed or are already being produced to support networking, mass storage, and so on. Pretty much anything you can buy a PCI card for today, you will be able to buy an Express card for now or in 1 to 2 years time.
There's even a PCMCIA-style form factor using PCI Express, and a daughterboard form factor for use in laptops.
comsider for example the time it takes to go over all possible flippings of 80 bits...
That's not what was asked. If a collision could be found by flipping exactly two bits, as grandparent wondered, then for an N-bit message you have only N*(N-1) combinations to try. So the correct answer to his question is simply "No (at least, not in the general case)". There are bound to be some messages out there -- very long and boring ones, for the most part -- that have that property.
A hash function is built exactly to prevent this kind of simple attacks.
What a sweeping generalization! In fact, the hash function implemented by XORing blocks of the message has *exactly* the property in question, for all messages.
Re:This reminds me of an old Styx tune ....
on
iPod Shuffle RAID
·
· Score: 1
Personally I feel more like the Rush lyrics: "Too many hands on my time".
A good solution is rotation of the backup media using a gray-code counter scheme. This means (conceptually) you keep a gray code counter of the number of backups performed, and the disk or tape you use for each backup corresponds to the bit that changes with the next increment of the counter. Example: if you have 4 DVD-RAMs in your rotation, labeled A through D, then you would rotate the media in the following sequence: ABACABADABACABAD... (no, it's not an extended version of a Genesis song). Now when you need to restore you have a wide range of ages of backups to choose from -- not just every incremental since the last full backup.
Let's say your hard disk crashes just after you've completed this much of the sequence:
ABACABADABA
Assuming your backups are daily, that means the backups in your stack have the following ages:
A: 0 days B: 1 day D: 3 days C: 7 days
Thus instead of being able to go back only 3 or 4 days, as with incremental backups using the same number of media, you can go back 7 days (there is some fluctuation in this number depending on where you are in the cycle). The advantage increases exponentially with the number of media in your rotation: 8 to 15 days with 5 disks, 16 to 31 days with 6 disks, etc. The gaps get larger as you go back but that's the tradeoff; if your file was corrupted 10 days ago you might have to choose disk E, 15 days old, so you've lost 15 days of work, but at least you were able to get back to a good version at all.
Now you can use rsync to do each backup and still have the advantage of a history that is at least as long as 2^(N-2) backup cycles.
If you were going to reply and say "oh, but I only do it every X weeks", well- you'll now loose weeks of work if you loose a file/drive.
*Not* if you lose a drive, in any case, because the last backup will always contain the most recent version of the drive before the crash. The only exception would be if the disk crash happens *during* the backup (and this shouldn't matter with rsync), but then if you only have a single backup copy you are a fool.
Incidentally, if you only have a single full backup with incrementals, you run exactly that risk because your full backup might have a medium error.
I believe the effect grandparent is describing is the same as that described in the very article you linked:
An extremely small hole, however, can produce significant diffraction effects which will result in a less clear image.
Granted, that probably takes a smaller opening that (what I imagine) cell phone cameras have, but the next sentence mentions an additional effect:
Additionally, as the diameter of the hole approaches the thickness of the material in which it is punched, significant vignetting at the edges of the image will result, as less light will reach these areas.
a) the link may be full-duplexed but the spindles on the other end are not.
Of course not, the spindles don't transfer data. Maybe you meant the drive heads are not full duplex, or maybe you mean that the spindles aren't synchronized. The first isn't an issue for the host adapter, which can always maintain a full duplex transfer on its link if it needs it, and for a disk-to-switch link it won't matter anyway because it is a point-to-point connection that won't be saturated by the data rates of a single disk. On the second, I think the effects of rotational latency are lost in a large, busy SAN due to queuing of packets at every stage of the network.
b) Very few applications have sustained bursts of reads and writes. Most have periods of sustained bursts of reads or writes.
A valid point. I was referring to the peak demand that will load the link down and force packets to wait, and with the exception of backbones you won't see it that much. Still, There are situations in which FC will use that full-duplex bandwidth opportunistically to keep packets flowing. If a write is in progress and another disk has read data or status to send, it can do so right away. If a read is in progress and the host adapter wants to send data or another command (read *or* write), it can do so right away. Not so on the SCSI bus.
As google has shown, you don't need a SAN to store a cache of the web. Many commodity PCs with simple IDE disks (lower latency) can do the job in parallel faster.
True, but the Google approach isn't for every enterprise. Google doesn't need to back up all that data, and the cache doesn't require SAN-level storage management capabilities. If they lose a disk, no problem, they just get the data back on the next crawl. I doubt that Google uses the same sort of network for their corporate administration and engineering development.
Actually, the FC rate is in multiples of 1.0625 Gb/s, so it is 2.125 Gb/s for 2GFC, but you need to divide by 10 instead of 8 to get the symbol rate, so 2GFC delivers 212.5 GB/s. Unlike parallel SCSI, though, Fibre Channel is full duplex, so with a good mix of reads and writes FC will move around 400 GB/s.
Parallel SCSI has higher overhead than FC. Arbitration, selection, and messages (which still use asynchronous transfers) are bandwidth killers. And if pSCSI has lower latency, it is because you are comparing different topologies -- loop or a switched SAN vs. a 3 meter bus (or is it 1.5 meters for 320? I forget). This is not to suggest that FC should be used for direct attachment of disks -- it would be blazingly fast, but it wouldn't be cost-effective.
1 & 2: Funny, Daihatsu don't actually bother to put the passenger space (headroom, legroom, hip room, and shoulder room measurements) into their specs, do they? Looks like a tight fit, judging from the photos -- that car would be a small compact or maybe a subcompact here. Anyway, with a full passenger load, Sirion holds 225 liters of cargo. That works out to just under 8 cubic feet, less than half of the Prius. Fine for a commute or buying groceries, but I'm not going to take a vacation or pick up any large packages in one of those things.
3. I guess I'd have to live there to appreciate what that means. They didn't get an exemption here (never heard of an EPA exemption) and the test results are what they are: 51 highway, 60 city. What the regulatory agencies allow car manufacturers to get away with over there doesn't concern me.
Sorry I won't be able to read the report, then. If it's based on UK vehicles it won't be much interest to me anyway, since we have different requirements for emissions control systems here in the US no doubt the results here would not be the same.
Searching with Google, it's hard to find any complete specs on the three models you like so much.
The latest reference I found to the Charade was for a 1992 model. It's about 2 and a half feet shorter, and 4 inches narrower, than the Prius.
I was able to locate some slightly more detailed specs on the Sirion, which is newer, and which has less cargo capacity than the Civic, and slightly more than half that of the Prius. I don't know what "rental class B" means in the UK but in my book this is a subcompact car.
And one of the articles I turned up on the Passo called it "Toyota's smallest compact car". I think we should keep our comparisons within the same class of automobile. Still, I'd like to see some evidence that at least one of these cars really does have "comparable or slightly smaller" cargo space relative to the Prius, instead of only half.
No, I haven't mixed UK and US gallons. All my figures are in US gallons -- the Prius's 51 MPG and the numbers I cited for the Civic. Even if you adjust the Civic's published US numbers to Imperial gallons it won't be 67 MPG -- actually around 60, but you'd have to adjust the Prius's numbers likewise. Maybe the British versions are rated higher than the US versions due to differences in emissions controls.
Toyota Prius does not list a highway spec because it is in the not-funny category.
Sorry, this comment is lost on me. Did Toyota decide their highway mileage for the Prius was not funny enough to publish, or did some reporter simply leave it out of their review? Toyota's website will give you the 51 MPG highway spec for the Prius as sold in the US.
The mileage estimates I used are from the EPA (the US Government's Environmental Protection Agency) tests, which were standardized sometime around 30 years ago. Those tests aren't exactly commensurate with reality, especially the city driving tests. All the same, a lot of Prius drivers do get results comparable to the EPA numbers. I concede that *I* don't, but my daily commute involves a lot of stop lights and a couple of big hills, which will affect the mileage of any vehicle (as a matter of physics). Highway mileage is different; at 65 MPH I get 45 to 50 miles per US gallon, close to that EPA number. The EPA figure represents an average speed of 48 MPG and a top speed of 60 MPG. Civic mileage at higher speeds would drop off from the EPA numbers faster than the Prius, due to Civic's higher aerodynamic drag.
I am referring to the much less publicised emission, efficiency and tax band rankings (UK, DE, a few others).
Either provide a link, or at the very the least, the title of the report.
[Me:] I'm not at all familiar with the three models you cited from Daihatsu and Toyota. Are they midsized (like the Prius), compact (like the Honda Civic), or subcompact (like the Honda Insight)? I just want to make sure it's a fair comparison.
[You:] (silence)
You didn't address this concern at all. I suspect you are comparing a midsized sedan to subcompact models. Even the Civic isn't a perfect comparison to the Prius since the Civic is a compact sedan. Still, it's the nearest-class competitor, so...
Double check the latest Civic specs. Honda has also updated the Civic engine train with the improvements that will go into the Hybrid Accord. It is just no as loud as Toyota about it. It now delivers 67+ mpg highway which is clearly better then the Toyota. In fact it is the only hybrid that approaches diesel as far as highway MPG are concerned.
No, youdouble-check. The fact sheet linked from that page states that the '05 Civic Hybrid (page 3) gets EPA results of 46/51 with the manual, 48/47 with the CVT. Your 67+ MPG is a made-up number. Also, here's a direct comparison between the two cars.
Also, as I pointed out in another comment, comparisons with diesel are not fair since diesel fuel has higher energy content than gasoline (130000 BTU per gallon, vs. 115000). Add about 13% to the gasoline car's MPG numbers before you start to compare apples and oranges on technical merits.
because Plancks constant is so 'small' as compared to say Avogadros number
Is Planck's constant really small, or is it just that the units it is expressed in are large?
More to the point, how is Avogadro's number a relevant reference for comparison? It's not a physical constant, anymore than "5280 feet per mile" is a physical constant; rather, it's nothing but a conversion factor.
Slashdot Mirror
One (sung by Gandalf, to the tune of "One"):
Well it's - one Ring,
To bind,
And hold them in the darkness -
Tonight
It's One, but we're, not the same
We'll just hurt each other
Somehow, I just can't hear those lyrics to Three Dog Night's music.
Binary $1,000,000.00 would be decimal $64.00. I think what you are looking for is binary 100000000 cents = $2.56.
That's right, one binary operand.
I don't know what definition of "clock skew" he is using but to say you can measure clock skew by looking at time stamps generated from a single point makes no sense at all. Clock skew is the mean time difference between *two* clocks of the same frequency (or related by a rational multiplier). The skew of any clock measured against itself at the same point in a clock tree is always zero.
It seems likely to me that he is measuring the differences in clock frequency from a reference frequency. This is not clock skew. Call it "clock frequency variance" or something to that effect.
So why do you typically get three Wilco songs in an hour while Aretha Franklin waits in the wings forever? Yep, mine plays "Less Than You Think" over and over.
OK, I've already made one pedantic comment today, but I'll risk another. There is definitely not another Mersenne prime between the 41st and 42nd Mersenne primes.
It has reached the point where proving a number prime is MUCH easier than finding any factors of it.
Unless the number in question is composite. In that case, it is MUCH easier to find factors of it, than to prove that it is a prime.
Perhaps it is a time-reversed galaxy. So, instead of emitting photons, it absorbs them.
I don't know if you meant to say something else, but you cannot "easily add more lanes to each slot". If the root complex or switch driving that slot has 4 lanes, 4 lanes is all you will get from that slot; if it has 8 lanes, 8 lanes is all you will get.
it's primarily a video interface on the desktop. No. It *is* replacing AGP, but it isn't limited to that. PCI Express devices are being developed or are already being produced to support networking, mass storage, and so on. Pretty much anything you can buy a PCI card for today, you will be able to buy an Express card for now or in 1 to 2 years time. There's even a PCMCIA-style form factor using PCI Express, and a daughterboard form factor for use in laptops.
Hey, the FCC is an arm of the people you helped to elect.
Er, no, it's an arm of the people I tried to defeat.
comsider for example the time it takes to go over all possible flippings of 80 bits ...
That's not what was asked. If a collision could be found by flipping exactly two bits, as grandparent wondered, then for an N-bit message you have only N*(N-1) combinations to try. So the correct answer to his question is simply "No (at least, not in the general case)". There are bound to be some messages out there -- very long and boring ones, for the most part -- that have that property.
A hash function is built exactly to prevent this kind of simple attacks.
What a sweeping generalization! In fact, the hash function implemented by XORing blocks of the message has *exactly* the property in question, for all messages.
Personally I feel more like the Rush lyrics: "Too many hands on my time".
Friday the 13th come on a Monday the 24th this year!
I heard New Line has 480,000 shields up for auction, that were used in the production of LotR. Sounds like a match to me!
A good solution is rotation of the backup media using a gray-code counter scheme. This means (conceptually) you keep a gray code counter of the number of backups performed, and the disk or tape you use for each backup corresponds to the bit that changes with the next increment of the counter. Example: if you have 4 DVD-RAMs in your rotation, labeled A through D, then you would rotate the media in the following sequence: ABACABADABACABAD... (no, it's not an extended version of a Genesis song). Now when you need to restore you have a wide range of ages of backups to choose from -- not just every incremental since the last full backup.
Let's say your hard disk crashes just after you've completed this much of the sequence:
ABACABADABA
Assuming your backups are daily, that means the backups in your stack have the following ages:
A: 0 days
B: 1 day
D: 3 days
C: 7 days
Thus instead of being able to go back only 3 or 4 days, as with incremental backups using the same number of media, you can go back 7 days (there is some fluctuation in this number depending on where you are in the cycle). The advantage increases exponentially with the number of media in your rotation: 8 to 15 days with 5 disks, 16 to 31 days with 6 disks, etc. The gaps get larger as you go back but that's the tradeoff; if your file was corrupted 10 days ago you might have to choose disk E, 15 days old, so you've lost 15 days of work, but at least you were able to get back to a good version at all.
Now you can use rsync to do each backup and still have the advantage of a history that is at least as long as 2^(N-2) backup cycles.
If you were going to reply and say "oh, but I only do it every X weeks", well- you'll now loose weeks of work if you loose a file/drive.
*Not* if you lose a drive, in any case, because the last backup will always contain the most recent version of the drive before the crash. The only exception would be if the disk crash happens *during* the backup (and this shouldn't matter with rsync), but then if you only have a single backup copy you are a fool.
Incidentally, if you only have a single full backup with incrementals, you run exactly that risk because your full backup might have a medium error.
According to the story, the dark side of the moon is almost perfectly bisected
There is no dark side of the moon, really.
As a matter of fact, it's all dark.
I believe the effect grandparent is describing is the same as that described in the very article you linked:
An extremely small hole, however, can produce significant diffraction effects which will result in a less clear image.
Granted, that probably takes a smaller opening that (what I imagine) cell phone cameras have, but the next sentence mentions an additional effect:
Additionally, as the diameter of the hole approaches the thickness of the material in which it is punched, significant vignetting at the edges of the image will result, as less light will reach these areas.
Why is encryption necessary on a product that the user must be able to read in the first place?
What's next, encrypted books, newspapers, and magazines?
a) the link may be full-duplexed but the spindles on the other end are not.
Of course not, the spindles don't transfer data. Maybe you meant the drive heads are not full duplex, or maybe you mean that the spindles aren't synchronized. The first isn't an issue for the host adapter, which can always maintain a full duplex transfer on its link if it needs it, and for a disk-to-switch link it won't matter anyway because it is a point-to-point connection that won't be saturated by the data rates of a single disk. On the second, I think the effects of rotational latency are lost in a large, busy SAN due to queuing of packets at every stage of the network.
b) Very few applications have sustained bursts of reads and writes. Most have periods of sustained bursts of reads or writes.
A valid point. I was referring to the peak demand that will load the link down and force packets to wait, and with the exception of backbones you won't see it that much. Still, There are situations in which FC will use that full-duplex bandwidth opportunistically to keep packets flowing. If a write is in progress and another disk has read data or status to send, it can do so right away. If a read is in progress and the host adapter wants to send data or another command (read *or* write), it can do so right away. Not so on the SCSI bus.
As google has shown, you don't need a SAN to store a cache of the web. Many commodity PCs with simple IDE disks (lower latency) can do the job in parallel faster.
True, but the Google approach isn't for every enterprise. Google doesn't need to back up all that data, and the cache doesn't require SAN-level storage management capabilities. If they lose a disk, no problem, they just get the data back on the next crawl. I doubt that Google uses the same sort of network for their corporate administration and engineering development.
Actually, the FC rate is in multiples of 1.0625 Gb/s, so it is 2.125 Gb/s for 2GFC, but you need to divide by 10 instead of 8 to get the symbol rate, so 2GFC delivers 212.5 GB/s. Unlike parallel SCSI, though, Fibre Channel is full duplex, so with a good mix of reads and writes FC will move around 400 GB/s.
Parallel SCSI has higher overhead than FC. Arbitration, selection, and messages (which still use asynchronous transfers) are bandwidth killers. And if pSCSI has lower latency, it is because you are comparing different topologies -- loop or a switched SAN vs. a 3 meter bus (or is it 1.5 meters for 320? I forget). This is not to suggest that FC should be used for direct attachment of disks -- it would be blazingly fast, but it wouldn't be cost-effective.
1 & 2: Funny, Daihatsu don't actually bother to put the passenger space (headroom, legroom, hip room, and shoulder room measurements) into their specs, do they? Looks like a tight fit, judging from the photos -- that car would be a small compact or maybe a subcompact here. Anyway, with a full passenger load, Sirion holds 225 liters of cargo. That works out to just under 8 cubic feet, less than half of the Prius. Fine for a commute or buying groceries, but I'm not going to take a vacation or pick up any large packages in one of those things.
3. I guess I'd have to live there to appreciate what that means. They didn't get an exemption here (never heard of an EPA exemption) and the test results are what they are: 51 highway, 60 city. What the regulatory agencies allow car manufacturers to get away with over there doesn't concern me.
Sorry I won't be able to read the report, then. If it's based on UK vehicles it won't be much interest to me anyway, since we have different requirements for emissions control systems here in the US no doubt the results here would not be the same.
Searching with Google, it's hard to find any complete specs on the three models you like so much.
The latest reference I found to the Charade was for a 1992 model. It's about 2 and a half feet shorter, and 4 inches narrower, than the Prius.
I was able to locate some slightly more detailed specs on the Sirion, which is newer, and which has less cargo capacity than the Civic, and slightly more than half that of the Prius. I don't know what "rental class B" means in the UK but in my book this is a subcompact car.
And one of the articles I turned up on the Passo called it "Toyota's smallest compact car". I think we should keep our comparisons within the same class of automobile. Still, I'd like to see some evidence that at least one of these cars really does have "comparable or slightly smaller" cargo space relative to the Prius, instead of only half.
No, I haven't mixed UK and US gallons. All my figures are in US gallons -- the Prius's 51 MPG and the numbers I cited for the Civic. Even if you adjust the Civic's published US numbers to Imperial gallons it won't be 67 MPG -- actually around 60, but you'd have to adjust the Prius's numbers likewise. Maybe the British versions are rated higher than the US versions due to differences in emissions controls.
Toyota Prius does not list a highway spec because it is in the not-funny category.
Sorry, this comment is lost on me. Did Toyota decide their highway mileage for the Prius was not funny enough to publish, or did some reporter simply leave it out of their review? Toyota's website will give you the 51 MPG highway spec for the Prius as sold in the US.
The mileage estimates I used are from the EPA (the US Government's Environmental Protection Agency) tests, which were standardized sometime around 30 years ago. Those tests aren't exactly commensurate with reality, especially the city driving tests. All the same, a lot of Prius drivers do get results comparable to the EPA numbers. I concede that *I* don't, but my daily commute involves a lot of stop lights and a couple of big hills, which will affect the mileage of any vehicle (as a matter of physics). Highway mileage is different; at 65 MPH I get 45 to 50 miles per US gallon, close to that EPA number. The EPA figure represents an average speed of 48 MPG and a top speed of 60 MPG. Civic mileage at higher speeds would drop off from the EPA numbers faster than the Prius, due to Civic's higher aerodynamic drag.
I am referring to the much less publicised emission, efficiency and tax band rankings (UK, DE, a few others).
Either provide a link, or at the very the least, the title of the report.
[Me:] I'm not at all familiar with the three models you cited from Daihatsu and Toyota. Are they midsized (like the Prius), compact (like the Honda Civic), or subcompact (like the Honda Insight)? I just want to make sure it's a fair comparison.
[You:] (silence)
You didn't address this concern at all. I suspect you are comparing a midsized sedan to subcompact models. Even the Civic isn't a perfect comparison to the Prius since the Civic is a compact sedan. Still, it's the nearest-class competitor, so...
Double check the latest Civic specs. Honda has also updated the Civic engine train with the improvements that will go into the Hybrid Accord. It is just no as loud as Toyota about it. It now delivers 67+ mpg highway which is clearly better then the Toyota. In fact it is the only hybrid that approaches diesel as far as highway MPG are concerned.
No, you double-check. The fact sheet linked from that page states that the '05 Civic Hybrid (page 3) gets EPA results of 46/51 with the manual, 48/47 with the CVT. Your 67+ MPG is a made-up number. Also, here's a direct comparison between the two cars.
Also, as I pointed out in another comment, comparisons with diesel are not fair since diesel fuel has higher energy content than gasoline (130000 BTU per gallon, vs. 115000). Add about 13% to the gasoline car's MPG numbers before you start to compare apples and oranges on technical merits.
Is Planck's constant really small, or is it just that the units it is expressed in are large?
More to the point, how is Avogadro's number a relevant reference for comparison? It's not a physical constant, anymore than "5280 feet per mile" is a physical constant; rather, it's nothing but a conversion factor.