That would indeed be tricky, though I've also thought about that as a music compression system. But this sounds more like RAID or PAR2, only with packets instead of disks/files.
They're measuring speed as "size of file / time it takes to get a usable copy". The size is going up some from redundant info, but the time to get a usable copy is dropping a lot because there's no waiting for retransmits; the delay before retransmitting is _huge_ compared to the actual data-shovelling time.
Here's an example. Say you're downloading 4.5 megabytes of data over TCP. That's 36 megabits, and on a 54megabit/sec link it should take 2/3 of a second. But that's also about 3000 packets (at about 1500 bytes each) and if you're losing 2% of them that's 60 packets. Now comes the brutal part: it takes the sender about a second to stop waiting for acknowledgement, decide that the packet didn't make it, and send a replacement. If the replacement gets lost the sender will wait twice as long before trying again, and so forth, but let's ignore that for now. So each of those 60 packets takes about a second to retry, and now instead of 2/3 of a second it takes 60 seconds, a factor of 90 slowdown.
of course there's a lot more details to it and the numbers are back-of-napkin grade, but that's why it's not 2%; 2% loss can result in much more than 2% slowdown because of how retries are handled.
well, first you'd have to find a place that has real competition, which to me means a situation where a new business could reasonably start up in competition with the incumbents. This might be difficult, because of barriers to entry; the US is probably not a good example now (though back when the incumbents were required to resell bandwidth to anyone who wanted it, it was closer, imho). It still might be better than many, but I'd be checking out Japan and/or Korea before I settle on the US.
there is a difference between "existed" and "became the norm". Until 1988 there were under 75k submissions of US origin per year, usually under 70k. Now it's more like 200-250k per year. I believe this indicates a shift in mindset that places a lot more emphasis on patents as a business tool than existed previously.
Forgot bandwidth fees. So figure that in 7 months you pay for half a drive and the bandwidth, and you're at 2.75 hours per day per user without dedupe. Tighter, but might still be feasible.
or they could just say the heck with it; 15.99 a month will buy a 2TB drive in 7 months. Figure 5 more months will buy some redundancy and a profit margin and they just have to expect folks won't record more than 2TB a year on average. Using DVD format you get about 2G/hour, so that's 1000 hours. Using a better compressor you can probably double that, which makes about 5.5 hours per day. That's probably a safe average, at least for now.
of course if they can come up with anything in the way of viable deduplication, it skyrockets, because if folks are considering it worth recording, probably lots of folks are recording it and the space usage per person drops like a rock.
that would be cool, because they either wouldn't keep commercials or would have commercials for a (probably) different region, so I can see something new:)
they are free to use whatever open source they can while building their solution. Delivering existing open source stuff as part of the solution may not work as well, depending on the exact license, and that may force the competition to redevelop the wheel, increasing their costs.
that would be the incredible breakthroughs. It's gonna take some serious improvements to stay even, much less keep total available energy per capita stable.
more the latter, I think. They're overworked and understaffed and their directive seems to be "pass it unless there's an obvious an egregious problem and let the courts sort out anything else".
as it happens the law does not restrict fair use to only those purposes, but uses them as examples. Given how common such things probably are on Youtube, getting a court ruling seems like a good idea. (Maybe if the automatic systems are required to get a human to agree, news stations will stop taking down NASA footage.)
I think you're underestimating the number of featuresets that the market will demand. At work, we really do kind of need the 5 trays of variously sized paper and the collator unit. At home I can use a 50-sheet tray that's built in to the printer, and I don't want to pay for the extra hardware that it would need to interface with the collator and extra trays, because at home it will never get to use it.
And while I like my little color laser at home for a lot of things, my inkjet does a much better job of printing on my DVD blanks, and I think it does a better job with my vacation photos too.
unless the prof requires the new edition, that switches the order of chapters 9 and 10, makes minor changes in the constants in the word problems throughout the book so the answers are different, and adds 10 more bucks to the price over last year's edition.
Slashdot Mirror
That would indeed be tricky, though I've also thought about that as a music compression system. But this sounds more like RAID or PAR2, only with packets instead of disks/files.
They're measuring speed as "size of file / time it takes to get a usable copy". The size is going up some from redundant info, but the time to get a usable copy is dropping a lot because there's no waiting for retransmits; the delay before retransmitting is _huge_ compared to the actual data-shovelling time.
Here's an example. Say you're downloading 4.5 megabytes of data over TCP. That's 36 megabits, and on a 54megabit/sec link it should take 2/3 of a second. But that's also about 3000 packets (at about 1500 bytes each) and if you're losing 2% of them that's 60 packets. Now comes the brutal part: it takes the sender about a second to stop waiting for acknowledgement, decide that the packet didn't make it, and send a replacement. If the replacement gets lost the sender will wait twice as long before trying again, and so forth, but let's ignore that for now. So each of those 60 packets takes about a second to retry, and now instead of 2/3 of a second it takes 60 seconds, a factor of 90 slowdown.
of course there's a lot more details to it and the numbers are back-of-napkin grade, but that's why it's not 2%; 2% loss can result in much more than 2% slowdown because of how retries are handled.
well, first you'd have to find a place that has real competition, which to me means a situation where a new business could reasonably start up in competition with the incumbents. This might be difficult, because of barriers to entry; the US is probably not a good example now (though back when the incumbents were required to resell bandwidth to anyone who wanted it, it was closer, imho). It still might be better than many, but I'd be checking out Japan and/or Korea before I settle on the US.
there is a difference between "existed" and "became the norm". Until 1988 there were under 75k submissions of US origin per year, usually under 70k. Now it's more like 200-250k per year. I believe this indicates a shift in mindset that places a lot more emphasis on patents as a business tool than existed previously.
Forgot bandwidth fees. So figure that in 7 months you pay for half a drive and the bandwidth, and you're at 2.75 hours per day per user without dedupe. Tighter, but might still be feasible.
or they could just say the heck with it; 15.99 a month will buy a 2TB drive in 7 months. Figure 5 more months will buy some redundancy and a profit margin and they just have to expect folks won't record more than 2TB a year on average. Using DVD format you get about 2G/hour, so that's 1000 hours. Using a better compressor you can probably double that, which makes about 5.5 hours per day. That's probably a safe average, at least for now.
of course if they can come up with anything in the way of viable deduplication, it skyrockets, because if folks are considering it worth recording, probably lots of folks are recording it and the space usage per person drops like a rock.
that would be cool, because they either wouldn't keep commercials or would have commercials for a (probably) different region, so I can see something new :)
practice makes perfect :)
they are free to use whatever open source they can while building their solution. Delivering existing open source stuff as part of the solution may not work as well, depending on the exact license, and that may force the competition to redevelop the wheel, increasing their costs.
that would be the incredible breakthroughs. It's gonna take some serious improvements to stay even, much less keep total available energy per capita stable.
In general looks I think I like the t-rex below, but the mclaren is better enclosed. I could go for this :)
I dunno. I like the trikes I've seen; if they added a windshell that would keep the rain off, I'd put it high on my list for my next car.
Depends on what you're willing to sacrifice to further the mainstream adoption of Linux on the desktop by non-technical users.
ftfy. A corporation has a profit motive to expand their user base. The linux kernel devs, not so much...
more the latter, I think. They're overworked and understaffed and their directive seems to be "pass it unless there's an obvious an egregious problem and let the courts sort out anything else".
that's how they will get away with not putting enough info in the patent to implement.
as it happens the law does not restrict fair use to only those purposes, but uses them as examples. Given how common such things probably are on Youtube, getting a court ruling seems like a good idea. (Maybe if the automatic systems are required to get a human to agree, news stations will stop taking down NASA footage.)
It's cheaper to just pretend it's a completely unrelated story that happens to have the same name, and ignore it. See also "Starship Troopers".
I think you're underestimating the number of featuresets that the market will demand. At work, we really do kind of need the 5 trays of variously sized paper and the collator unit. At home I can use a 50-sheet tray that's built in to the printer, and I don't want to pay for the extra hardware that it would need to interface with the collator and extra trays, because at home it will never get to use it.
And while I like my little color laser at home for a lot of things, my inkjet does a much better job of printing on my DVD blanks, and I think it does a better job with my vacation photos too.
unless the prof requires the new edition, that switches the order of chapters 9 and 10, makes minor changes in the constants in the word problems throughout the book so the answers are different, and adds 10 more bucks to the price over last year's edition.
a prop plane may not be fast enough to defend against jets, but it's fast enough to attack infantry or field installations (or probably cavalry)
It's a sine of the times that nobody reads TFGP
Since you disagree with this usage of the word, would you care to present an alternative definition for peer review?
It doesn't really affect software, as far as I can tell.
You're allowed to put up projects for things that don't exist. You're not allowed to make it look like it does exist already.
It prohibits things like the 'reseller kit' levels, where you get multiple of the item and a display case.