It turns out that a typical HTTP/1.1 request requires multiple TCP packets to get all the headers across. With TCP slow start, this takes a long time because only one packet gets transmitted per round trip in the beginning. Obviously this gets even worse if you try to browse to a different continent, with 100ms+ latency.
HTTP/2 manages to fit most requests into one packet, assuming a reasonable MTU. To do this requires both a binary protocol encoding and header compression. Without those, you need two packets which is half as fast.
Of course you could argue that this is all because TCP is a stupid ancient protocol which no one sane should be using in 2014.
It is difficult to imagine a destroyed Earth that is less hospitable than Mars. Not impossible, but difficult. In almost all cases it is easier to terraform Earth than Mars.
It would even be easier to build deep underwater communities on Earth.They are unlikely to be destroyed by climate change or ecosystem collapse, and resources are vastly easier to get there.
One of the challenges is that there is just so much data from modern experiments. E.g. the LHC throws a lot of irrelevant data away before it even hits the hard drives. To a layman like me it seems likely that some potentially useful data gets thrown away in this process.
You cannot fault the LHC designers for doing this; the data handling and storage there is awe-inducing. Transporting and storing orders of magnitude of (probably useless) more data is just not feasible.
That is the trick: You do not need to pump water anywhere. You just avoid using water when other forms of energy are delivering. If there is a power surplus even with the turbines off, you try to export that somewhere, and if no one wants it even for free, you stop the wind turbines or solar plants or whatever. If you happen to have pumped storage available for the few hours a year with low or negative prices, great, but otherwise it is not a great loss.
The neat thing about reservoir-based hydro power is that you can multiply its output by using pretty much any intermittent power source. If the rainfall is cut in half, you "just" build enough wind turbines or solar cells or whichever to provide the missing half of yearly energy output, and save up water whenever you can. At a large enough scale, you can boost hydro power many times over, assuming the hydro generators themselves are large enough. Luckily hydro generators are reasonably cheap to upgrade, whereas reservoir capacity is hard to come by.
Why would any bank finance a car loan without insurance? That would be monumentally stupid.
The bank can just insure the car themselves and add the cost to the interest rate they charge. However, most banks are likely large enough that self-insurance is the best option for them.
3D TV requires you to disassociate your depth perception and your eye focus. 3D needs to stay properly sharp despite your eyes changing focus (and ideally blur the things which are out-of-focus, but that is less important). The technology to do so is almost here now.
Alternatively, most children today probably watch enough 3D that their vision adapts.
We have way more gold in storage than we have an industrial or jewelry-related use for. The primary use for gold is to sit in Fort Knox doing nothing at all. Gold-coloured plastic would do that job just as well.
Preventing Russia from placing drilling platforms in Canadian waters is a use. If your enemy has air superiority, all you can do is look on angrily like Vietnam does.
1080i is a disaster for sports. Then again, so are all the other things cable companies do to ruin the signal.
At least there is finally hope: 4k has no interlacing, and with 8x8 *PEG blocks you still get the equivalent of 512x270 uncompressed pixels, no matter how crappy the encoding is. That is nearly NTSC resolution...
First off, it seems unlikely as the phone will either be a corporate device, including BYOD, or a personal device. In the first case the traffic will have to flow across the network (including the firewall)
What stops a BYOD from using multipath? It will have to use the 4G connection when it isn't on the corporate wifi, so what keeps it from using both?
USB device drivers are not of sufficient quality to make that mitigation very viable. Just exploit the broken drivers instead; on most operating systems device drivers have the equivalent of root privileges.
Wrong way around surely: The test audience found the movie confusing and sad and so the internal monologue and the happy ending was added. Later came the director's cut which attempted (unsuccessfully) to outdo 2001: A Space Odyssey for longest CGI scene with nothing happening.
The reason why routers are so underpowered is that nobody uses multicast. If there was a strong demand for multicast I'm sure that the manufacturers would increase the capacity of their hardware.
If you build it and it costs less than 6 figures USD, you will drown in customers. It would not be used primarily for multicast at first, it would be used to get BGP working better, but every major ISP would want your router.
Using P2P does not lower the total load on the network, it just spreads it out more evenly.
Correctly done P2P sends traffic through the best route, typically from someone on the same ISP as the recipient and preferably from the same neighbourhood. That lowers total load a lot. Most current P2P networks do not particularly worry about optimal routing; they are much more constrained by traffic shaping or (often artificially) limited last-mile upstream capacity. It would be fairly easy to give priority to low-latency peers.
Besides, P2P can solve the problem of subscribers not watching at exactly the same time. Multicast breaks as soon as someone presses pause.
Multicast is not a viable technology for truly large scale deployments (more than a few hundred thousand hosts perhaps). Routers and switches do not have the required resources to maintain multicast routing/switching tables for millions of multicast sessions.
The correct way to solve the problem is to push it to the end nodes. They have much more CPU power and memory than routers and switches. The technology to do so has existed for a long time: P2P.
Thats how the internet is paid for. The sending provider pays the receiving provider for the bandwidth, and this is the only rational way it can be.
No. That is not how it works. The truth is that the smaller provider pays the larger provider, no matter which direction the traffic flows. Some companies, like Netflix, are nice enough to not use their size as an excuse to charge people -- they offer free peering at internet exchanges. Other companies are maximally greedy.
Slashdot Mirror
It is actually surprisingly complicated.
It turns out that a typical HTTP/1.1 request requires multiple TCP packets to get all the headers across. With TCP slow start, this takes a long time because only one packet gets transmitted per round trip in the beginning. Obviously this gets even worse if you try to browse to a different continent, with 100ms+ latency.
HTTP/2 manages to fit most requests into one packet, assuming a reasonable MTU. To do this requires both a binary protocol encoding and header compression. Without those, you need two packets which is half as fast.
Of course you could argue that this is all because TCP is a stupid ancient protocol which no one sane should be using in 2014.
It is difficult to imagine a destroyed Earth that is less hospitable than Mars. Not impossible, but difficult. In almost all cases it is easier to terraform Earth than Mars.
It would even be easier to build deep underwater communities on Earth.They are unlikely to be destroyed by climate change or ecosystem collapse, and resources are vastly easier to get there.
Anders Behring Breivik. Alas, he also had the skills to obtain guns legally.
Yes, the typical use case for a MacBook is serving bash scripts over HTTP. Patch quickly!
One of the challenges is that there is just so much data from modern experiments. E.g. the LHC throws a lot of irrelevant data away before it even hits the hard drives. To a layman like me it seems likely that some potentially useful data gets thrown away in this process.
You cannot fault the LHC designers for doing this; the data handling and storage there is awe-inducing. Transporting and storing orders of magnitude of (probably useless) more data is just not feasible.
That is the trick: You do not need to pump water anywhere. You just avoid using water when other forms of energy are delivering. If there is a power surplus even with the turbines off, you try to export that somewhere, and if no one wants it even for free, you stop the wind turbines or solar plants or whatever. If you happen to have pumped storage available for the few hours a year with low or negative prices, great, but otherwise it is not a great loss.
The neat thing about reservoir-based hydro power is that you can multiply its output by using pretty much any intermittent power source. If the rainfall is cut in half, you "just" build enough wind turbines or solar cells or whichever to provide the missing half of yearly energy output, and save up water whenever you can. At a large enough scale, you can boost hydro power many times over, assuming the hydro generators themselves are large enough. Luckily hydro generators are reasonably cheap to upgrade, whereas reservoir capacity is hard to come by.
Why would any bank finance a car loan without insurance? That would be monumentally stupid.
The bank can just insure the car themselves and add the cost to the interest rate they charge. However, most banks are likely large enough that self-insurance is the best option for them.
3D TV requires you to disassociate your depth perception and your eye focus. 3D needs to stay properly sharp despite your eyes changing focus (and ideally blur the things which are out-of-focus, but that is less important). The technology to do so is almost here now.
Alternatively, most children today probably watch enough 3D that their vision adapts.
Procurve is getting killed by software development. They are behind their competitors, and getting left further behind each day.
If you can make do with the limited software features (and you avoid the rebranded 3com stuff) they are great switches.
We have way more gold in storage than we have an industrial or jewelry-related use for. The primary use for gold is to sit in Fort Knox doing nothing at all. Gold-coloured plastic would do that job just as well.
There are plenty of places that run HD over cable on plain -C or even -T.
DRM does not have any connection to which particular DVB version is used.
Preventing Russia from placing drilling platforms in Canadian waters is a use. If your enemy has air superiority, all you can do is look on angrily like Vietnam does.
1080i is a disaster for sports. Then again, so are all the other things cable companies do to ruin the signal.
At least there is finally hope: 4k has no interlacing, and with 8x8 *PEG blocks you still get the equivalent of 512x270 uncompressed pixels, no matter how crappy the encoding is. That is nearly NTSC resolution...
"+"-characters keep out an amazing amount of spam. Please do not teach the world to recognize them.
First off, it seems unlikely as the phone will either be a corporate device, including BYOD, or a personal device. In the first case the traffic will have to flow across the network (including the firewall)
What stops a BYOD from using multipath? It will have to use the 4G connection when it isn't on the corporate wifi, so what keeps it from using both?
Your IDS/IPS cannot look inside SSL traffic
Sure it can. You just push a new root certificate to your devices and intercept away.
USB device drivers are not of sufficient quality to make that mitigation very viable. Just exploit the broken drivers instead; on most operating systems device drivers have the equivalent of root privileges.
Personally I am much more excited about a gigatesla than a Tesla Giga.
Wrong way around surely: The test audience found the movie confusing and sad and so the internal monologue and the happy ending was added. Later came the director's cut which attempted (unsuccessfully) to outdo 2001: A Space Odyssey for longest CGI scene with nothing happening.
The reason why routers are so underpowered is that nobody uses multicast. If there was a strong demand for multicast I'm sure that the manufacturers would increase the capacity of their hardware.
If you build it and it costs less than 6 figures USD, you will drown in customers. It would not be used primarily for multicast at first, it would be used to get BGP working better, but every major ISP would want your router.
Using P2P does not lower the total load on the network, it just spreads it out more evenly.
Correctly done P2P sends traffic through the best route, typically from someone on the same ISP as the recipient and preferably from the same neighbourhood. That lowers total load a lot. Most current P2P networks do not particularly worry about optimal routing; they are much more constrained by traffic shaping or (often artificially) limited last-mile upstream capacity. It would be fairly easy to give priority to low-latency peers.
Besides, P2P can solve the problem of subscribers not watching at exactly the same time. Multicast breaks as soon as someone presses pause.
Multicast is not a viable technology for truly large scale deployments (more than a few hundred thousand hosts perhaps). Routers and switches do not have the required resources to maintain multicast routing/switching tables for millions of multicast sessions.
The correct way to solve the problem is to push it to the end nodes. They have much more CPU power and memory than routers and switches. The technology to do so has existed for a long time: P2P.
Thats how the internet is paid for. The sending provider pays the receiving provider for the bandwidth, and this is the only rational way it can be.
No. That is not how it works. The truth is that the smaller provider pays the larger provider, no matter which direction the traffic flows. Some companies, like Netflix, are nice enough to not use their size as an excuse to charge people -- they offer free peering at internet exchanges. Other companies are maximally greedy.
Netflix limited the highest bandwidth streams to direct peers for a while. Everyone going through transit could not get the highest quality.
Similarly, it would be quite trivial for Netflix to limit non-paying ISPs to lower quality.
We are talking about ODF here. The encoding will be UTF-8.