Vint Cerf Says Fix the Net With More Pipe 341
CWmike writes "While ISPs may fret about Netflix, Hulu and other streaming media services saturating their bandwidth, Internet forefather Vint Cerf has a simple answer for this potential problem: Increase bandwidth exponentially. With sufficient bandwidth, streaming video services of prerecorded content wouldn't be necessary, said the now-technology evangelist at Google. With sufficient throughput, the entire file of a movie or television show could be downloaded in a fraction of the time that it would take to stream the content. Cerf, speaking at Juniper Network's Nextwork conference, spoke about the company's decision to outfit Kansas City with fiber-optic connections that Google claims will be 100 times faster than today's services. The purpose of the project was 'to demonstrate what happens when you have gigabit speeds available,' Cerf said. 'Some pretty dramatic applications are possible.' One obvious application is greater access to high-definition video, he explained. 'When you are watching video today, streaming is a very common practice. At gigabit speeds, a video file [can be transferred] faster than you can watch it,' he said. 'So rather than [receiving] the bits out in a synchronous way, instead you could download the hour's worth of video in 15 seconds and watch it at your leisure.' He adds: 'It actually puts less stress on the network to have the higher speed of operation.'"
You heard the man! (Score:5, Funny)
Give her more pipe!
The internet... (Score:2)
Re: (Score:2)
and it's also made of cats
Both Dead and Alive?
Re: (Score:2)
Makes sense... (Score:5, Insightful)
'So rather than [receiving] the bits out in a synchronous way, instead you could download the hour's worth of video in 15 seconds and watch it at your leisure.' He adds: 'It actually puts less stress on the network to have the higher speed of operation.'
Sure, it naturally would stand to reason that the operations (like streaming video) that currently require 100% utilization on today's network might only require a fraction of that on tomorrow's much faster network. The problem is, tomorrow we won't be happy with the same old video we used to stream, we are going to want a super high-def version with 8 channel stereo sound and in-line twitter commentary plus it will have to update our facebook status every time we pause it to go to the bathroom... And then we will be back to streaming at 100% capacity again, wondering when the next leap in networking will let us do block downloads again.
Seriously, Vint Cerf? This is the best idea you can muster? This is the same problem/solution cycle the internet has been locked in for its entire existence.
Re: (Score:2)
The problem is, tomorrow we won't be happy with the same old video we used to stream
True. In the future, we'll actually want live news and live sports, the two areas where subscription Internet video has lagged behind cable and satellite TV.
Re: (Score:2)
That is a issue with the companies that own that content, not with the technology behind it. NHL.com is a great example, I thought it could get me hockey games but it only shows out of market. So instead of the NHL getting some of my money, they get none. I hope they like that arrangement.
How does live news lag online?
Websites update faster than the local talking head can keep up.
Can't surf while ironing or washing dishes (Score:2)
Websites update faster than the local talking head can keep up.
While doing housework, it's far more difficult to use a website than to use MSNBC.
Re: (Score:2)
So your real complaint is lack of audio/video presentation of news via the Internet?
Seems like MSNBC or anyone could provide that. Newshour is available online.
Again, this is more a problem of people for some reason not wanting my money. You can lead a horse to water, but this horse seems to not want to drink.
Re: (Score:2)
You do housework staring at a TV?
You must SUCK at housework.
Otherwise there are hundreds of live news audio streams on the internet... From liberal sex fiends to right wing nutjobs ot even Libertarians... It's all over the internet... did you even bother to look?
Re: (Score:2)
it will never happen. The sports conglomerates live blacking out games. and it's easier to circumvent a online blackout than it is on your TV. It's all about greed, nothing more.
Re:Makes sense... (Score:4, Interesting)
Re:Makes sense... (Score:5, Funny)
Some people...
Re:Makes sense... (Score:4, Funny)
Re:Makes sense... (Score:5, Insightful)
'So rather than [receiving] the bits out in a synchronous way, instead you could download the hour's worth of video in 15 seconds and watch it at your leisure.'
You mean actually have the file stored on your PC? OH, yeah, that'll go down well with the MAFIAA.
Re: (Score:3)
Then the ISPs just throttle most P2P connections except to/fro their "super peers" and internal traffic. Hardly anyone would complain as long as their downloads are still fast.
Re: (Score:3)
Re: (Score:2)
FYI, this is how Apple's rentals/"streaming" work. They may refer to it as streaming, but it's not. You get a copy of the file, and it's DRMed with an expiration. You do not have to wait for the whole file to download to being watching, but that doesn't make it the same as streaming. Heck, once you've got the whole file, you can watch offline.
It's also how XBox Live video rentals used to work before that got corrupted with all the "Zune Marketplace" crap. Again, once you got the whole file, you could w
Re: (Score:2)
Youtube videos also work this way. The video comes in faster than it is displayed and is buffered into a temp file (and there are ways to find and save this temp file as long as the browser isn't closed or the URL is changed).
Re: (Score:2)
Re: (Score:2)
I wonder how much bandwidth netflix could save by cutting the credits and intro music off tv shows they stream.
Re: (Score:2)
intro music on most shows is almost nonexistent now, and in movies it can be an integral part
Re: (Score:2)
All it takes is for someone to go over to someone else's house and watch a movie from a Bluray (which the MAFIAA is trying to sell) or pirated 1080p x264 Matroska file, and suddenly the streaming user doesn't want their streaming service anymore.
Wut? I have a blu-ray player. I have blu-rays. I also am happy with DVDs and SD streaming for some films. Depends on the film, and the price of streaming vs the price of buying.
Re: (Score:2)
Umlimited resolutions are just plain useless, remember, even your *eyes'* resolution is limited. I doubt 8 sound channels take that much bandwith but hardly anyone has the necessary setup to benefit from this anyway.
Re: (Score:2)
Re:Makes sense... (Score:5, Informative)
I don't think you've caught the point. It isn't merely that 100% of the network won't be required if we are faster. If the connection is fast enough, the video doesn't have to be streamed through difficult to throttle udp but instead can be transferred as a network friendly tcp transfer. UDP video transfer is a dirty hack implemented because it was the only way to get video of watchable quality through. We are no longer in the days of choppy unwatchable video on the internet and if we move away from dirty hacks like udp streaming I doubt anyone would go back to it.
Re: (Score:2)
UDP streaming is not only a dirty hack, it also uses more bandwidth than it actually needs. Every packet has bits of both the preceeding and following packets so that if a packet gets dropped or delayed or whatever, there is still enough information to not miss the "missing" information. By eliminating the need for redundant information in every packet, going TCP over high speed networks will LOWER the actual bits needed to be transmitted, reducing the congestion that is killing Netflix and other such servi
Re: (Score:2)
Congestion is killing netflix?
That is news to me, I don't even have cable anymore. Looks fine over my 25/25 connection.
Re: (Score:2)
Um, if you say so. There's nothing in UDP that does that, and unconditionally tripling the payload seems like a pretty brutal kludge. I'm not saying that typical video streaming doesn't work that way, just that I'd be surprised if that's that awful.
If it does work that way then, urgh, I completely agree that TC
Re: (Score:2)
It isn't UDP protocol, it is the "streaming" part that does that to provide "smooth" playback. Dropping data requires some level of error correction to ensure usable playback of a stream in UDP. That error correction is always in the form of extra data. How much is dictated by the streaming protocol (not layer 3 or 4) . UDP, unlike TCP doesn't have much(if any) error correction built into the protocol. TCP at least has Packet sequencing and can detect missing packets and request resend.
Re: (Score:2, Informative)
through difficult to throttle udp
I haven't had any difficulty throttling any sort of IP stream in almost 15 years. UDP, TCP, ICMP, IGMP, RAW (otherwise unknown payloads), you name it. Just because your little OS or linksys router doesn't do it doesn't mean real network equipment doesn't. Literally 15 years ago, throttling UDP to specific rates with no problem at all.
UDP video transfer is a dirty hack implemented because it was the only way to get video of watchable quality through.
UDP is used because a lost packet doesn't stop the stream, missed packet is just a missed packet. If its a miss on a small portion of a moving image, you probably won't not
Re:Makes sense... (Score:5, Insightful)
The problem is, tomorrow we won't be happy with the same old video we used to stream, we are going to want a super high-def version with 8 channel stereo sound and in-line twitter commentary plus it will have to update our facebook status every time we pause it to go to the bathroom...
Nobody is making "super high def" content, nor can existing display devices do "super high def."
.. uh, what? stereo is 2 channel sound. Didnt you know that when you start talking about channels, you negate the whole stereo thing? Also, audio hasnt been an issue for years... even the WORST broadband connections can stream UNCOMPRESSED audio in realtime, and a 15:1 LOSSLESS audio compression is a typical reality.
This entire argument is based on the fallacy that bandwidth needs will grow forever. Its simply not true. Prior to double-digit megabit connections, there was always media that couldnt be delivered in real-time.. but now there simply isnt any media that cannot be delivered in real-time on 10+ mbit connections, and that includes 3D HD video.
I realize that in your imaginary world, the bandwidth of content grows exponentially.. but thats just your imagination. The jump from SD to HD was not an exponential growth in the size of content.
As far as "8 channel stereo sound"
As far as twitter and facebook.. you are further proving that you have absolutely no fucking idea what you are taking about.
The only way current high end bandwidth will be insufficient is if there is a new media paradigm.. holographic (real 3D) media and so forth.. that'll be possible in 10 years, or so they have been saying for the last 60 years.
Re: (Score:2)
... ...
Nobody is making "super high def" content, nor can existing display devices do "super high def."
Digital movies are filmed in "super high def".
The fact that display technology has regressed since the advent of 1080p LCD TV's is not lost on me. 10 years ago I had a CRT monitor that could handle twice that resolution.
Re:Makes sense... (Score:4, Insightful)
Believe it or not, he is right on the money (figuratively speaking). What he is suggesting is the correct response to "net neutrality" laws. Myopic and purely profit-driven ISPs won't give their users what the users want, just the absolute minimum to make maximum profit. OTOH, investment in infrastructure makes "net neut" irrelevant, but obviously takes money.
Moreover, I think you are wrong to say better quality video will fill the pipes because 1080p video is more than we'll need for a while, and I bet in time we'll even get better compression algorithms to bring the filesizes down further. What will push the network is increased internet penetration, but we'll have to deal with that anyway...
Re: (Score:2)
Myopic and purely profit-driven ISPs won't give their users what the users want
That is silly. Of course ISPs will give their users what they want. They will not, however, lose money in the process of giving it to them.
And investment in infrastructure will not make "network neutrality" irrelevant, because neutrality is about treating traffic equally. It has nothing to do with giving users unlimited bandwidth for a low fixed price.
Re: (Score:3)
That is silly. Of course ISPs will give their users what they want. They will not, however, lose money in the process of giving it to them.
If what you say is true, then why is google bothering to roll out 1gb fiber in Kansas City -- which already has Comcast and Time Warner giving their customers "what they want" at what you imply is a reasonable profit margin?
Re: (Score:2)
You do realize current video technology (HD broadcast TV) is higher definition that most people can see in their homes right?
The point to that being, one or two resolution jumps and we're likely to be at the final point for TV. You don't need more pixels once the eye stops perceiving any change in the image. Yes, we'll still have morons like people who call themselves audiophiles and listen to 'high bitrate mp3s' but you can safely ignore those morons with TV just like you do with music now.
As far as bein
Re: (Score:2)
Re: (Score:2)
Luckally we are reaching our biological limit of our bandwidth needs. Lets say 2 80" 1:1 at 1000PPI Display Streaming at 120 FPS 32bit color. with 32 channel stereo 128bit sound, uncompressed per person.
So a 3tbs bandwith per person should be more then enough for anyone, for home use.
Re: (Score:2)
"...instead you could download the hour's worth of video in 15 seconds..."
Download? Download?
A working copy on your harddisk?
The content MAFIAA just got a heart attack.
Re: (Score:2)
Actually, most of that is really round-off error compared to 1080p, HD video.
Adding things like a Facebook updates, Twitter streams, a couple of extra audio channels and whatnot don't really add that much to the bandwidth utilization.
Getting people to purchase new video devices that can handle greater than 1080p video is going to take probably decades. We're still seeing a lot of people with standard-def (480i) and even half-HD (720p) sets -- with little to no inclination to upgrade.
Then all the production
Re: (Score:2)
The problem is, tomorrow we won't be happy with the same old video we used to stream, we are going to want a super high-def version with 8 channel stereo sound and in-line twitter commentary plus it will have to update our facebook status every time we pause it to go to the bathroom... And then we will be back to streaming at 100% capacity again, wondering when the next leap in networking will let us do block downloads again.
I think you're wrong. The human audio visual system has limited resolution/perception ability. Once something is "good enough" most people stick with it. They don't keep pushing. Once audio/video is free of perceptible noise and distortion and is clear and sharp most people are happy.
Consumer interest in HD audio formats like Super Audio CD... virtually zero. 16 bits at 44100 samples per second is enough for most people. Lossless formats like FLAC and Apple's lossless are here and yet most people stil
Re: (Score:2)
On the same note: you can already have this.
an automated DVD ripping pc like I set up with linux and handbrake CLI it's 15 seconds of my personal time to rip the DVD and I watch it at my leisure. Granted ADHD nuts will point out that it takes longer than 15 seconds because they have to stand there watching it rip, but normal people can go on doing other things while the computer does it on it's own.....
It just takes more effort to set it up, disney owned releases are not automatic, and you are acting as
Re: (Score:2)
Can you imagine if I had to stream it down every single time? Oh lord, I can see my cable internet provider cringing at the thought...
Re:Makes sense... (Score:5, Interesting)
Do you? No, I don't think you do, not with existing video techniques. If block downloads were of a format that used end to end compression instead of stream compression, you could gain additional savings by requiring the user to download the whole file. As it is, prerecorded video streams already buffer enough to fill the MTU on the network, meaning that any ability to transfer it faster will still result in the same net amount of bytes going across the network. Wake me up when we have practical end-to-end compression techniques for video...
Re: (Score:2)
What good reason is there to stream recorded media in real time? It is not live-as-it-happens. It's like buying a film movie frame by frame and watching it as you receive the frames from the store. Makes absolutely no sense.
What if you want to watch a movie without waiting 30 minutes for the download to complete first?
Re: (Score:3)
So the inaccuracy is that you disagree with what I'm saying. Nice. Good luck getting past a 0 mod. Don't let a little thing like actually having a point to make stop you from being a raging asshole. FYI, not everyone has a connection that can handle 6-10 HD Netflix streams, and unless you are paying for 25Mbit+ service, neither do you. And my comment about MTU was for the exact reason you pointed out; the only time that efficiency goes down when streaming is if the MTU is unmet and a packet goes across
Re: (Score:3)
I think he's talking about the overhead of streaming, and making the point that once you have the speed to download rather than stream, you free up even more bandwidth than you would expect.
Of course what he doesn't talk about is how scared content providers will be when they hear that the user will have downloaded the file. Downloaded?!?!? Oh, noes - piracy! Arrgh!!
I'm embarrassed for /. that it took an AC to make this point. THIS is the key point in the conversation, and the reason Cerf is living in fantasyland. As long as the service providers are the same people with the pipe monopoly, they will continue to gobble up the content providers and NEVER support building out networks further. At every step of the way it is against their sense of self-interest, and it will be a major factor in the decline of US competitiveness over the next 30-40 years.
The public good is
Happy Birthday Vint! (Score:4, Informative)
(It's today.)
Re: (Score:3)
Same as Alan Turing's?
Re: (Score:2)
I guess so. Had made the connection before now.
Only one problem (Score:2)
Re: (Score:2)
Parent is correct. Streaming is the most user-accepted form of DRM out there.
Re: (Score:2)
Absolutely. It's glaringly obvious that 'buffering' without restriction is best for network utilization, but the provider throttles the connection rate for various reasons (memory consumption vs. disk space is the most technical one), but the biggest thing is 'rental' model is a little more palatable as a streaming implementation that *expressly* prevents faster-than-realtime download out of paranoia of piracy.
What about devices w/ insufficient local storage? (Score:3, Insightful)
The Wii won't be able to hold an entire downloaded movie --- unless one makes putting in a blank 8GB SD card before watching --- I don't think that will go over well, nor do I think the copyright holders will like the idea of a single monolithic file being made available.
The problem isn't merely a technical one...
William
Re: (Score:3)
It may not DL an entire movie, but it could grab a significant chunk of it and let you watch it without jitter or pixelation, and then DL the next chunk when you have a few minutes of video from the previous chunk.
The copyright holders are going to whine and moan no matter how it's done, so the best thing is to ignore them and do it the right way. It's just as easy to capture a streamed movie as it is one that arrive as a single piece. The thing they need to get over is that once people can rely on having a
Re: (Score:2)
The Wii won't be able to hold an entire downloaded movie
That's easily solved too. Increase storage... exponentially!
Re: (Score:2)
Re: (Score:2)
When the bill comes due... (Score:2)
I'm Trent the Thief, and I approve this message (Score:3)
Listen to the man, he knows of that which he speaks.
More bandwidth may not solve all the problems, but it'll sure as hell solve some of them.
I don't know how many of you still remember the dialup days, or even used dial-up. When the schoolkids got home, they'd start hitting AOL and you'd notice the lag.
It's not as bad now, what with me having a 25/25Mbps line. But there's still a very wide range of criminal acts that I'd perform to have my own 1 Gbps line.
Re: (Score:2)
Listen to the man, he knows of that which he speaks.
More bandwidth may not solve all the problems, but it'll sure as hell solve some of them.
I don't know how many of you still remember the dialup days, or even used dial-up. When the schoolkids got home, they'd start hitting AOL and you'd notice the lag.
It's not as bad now, what with me having a 25/25Mbps line. But there's still a very wide range of criminal acts that I'd perform to have my own 1 Gbps line.
Of course, most parts of the country have 1.5mbs or less for their internet connection. Should the emphasis be on providing a few people, in select metropolitan areas, unbelievable bandwidth or should it be on providing reasonable bandwidth for the rest of the country?
Re: (Score:2)
The two aren't mutually exclusive, but you have to get an order of magnitude improvement somewhere to generate an incentive to roll it out elsewhere. A gigabit MAN link three years ago was $100,000/month in my area. Today it can be had at the same location for about $5-7,000. While my little company didn't go for that much bandwidth, a 100M link is within reason and makes all kinds of interesting things possible. If there is a market for 100M links (at ~$2k/month, roughly what you would have paid for a
never happen (Score:5, Interesting)
unfortunately there is no way this will happen. There are too many important competing interests which act at the beaurocratic/governance level which are anti-bandwidth.
MPAA/RIAA don't want people to stream quickly because they fear content being stolen
CIA/FBI don't want increased bandwidth because they need(or think they need) to be able to monitor and index all communication (TIA)
ATT/Verizon and other telecoms don't want to because it represents a cost that will interfere with their milking of customers
Comcast doesn't want it because it will interfere with their control over content
Everyone just wants to stay status quo or worse. This will never happen.
Just wave that magic wand (Score:2)
I work for a small ISP in a mid-sized city, and I'd love to see more bandwidth everywhere. However, we can't get gigabit access for our company for a practical cost, much less deliver it to end users. There are real costs involved in running a physical plant that just don't scale up at this time.
In any case, the "download an hour of video in 15 seconds" is somewhat impractical in any case; downloading an hours worth of anything in 15 seconds requires 240 times the bandwidth over streaming it. Over-the-ai
Re: (Score:2)
Over-the-air HD video is up to 19 megabits per second, so the equivalent download would require a 4.6 gigabit/second link (at the end-user side; the server side would have to be many times that).
Dont confuse the inefficiencies of an implementation (an MPEG2 encoding) with a limitation in reality. 6 to 8 mbps is quite often more than fine for 1080p with H.264 (same error/pixel as DVD)
Re: (Score:2)
The point is that running a line at full capacity might be an efficient use of that particular infrastructure, but it doesn't share well. A single gigabit link upstream can serve a lot of customers with a gigabit handoff, until you actually have applications that can saturate that provision.
Re: (Score:2)
You're ignoring the overhead imposed by streaming protocols. The total bandwidth consumed by streaming is higher than that consumed by delivering the same bits via TCP. Of course, you can stream over TCP, but only if you have a large local buffer and only if your connection is enough faster than the minimum bitrate needed so that you can fill that buffer and keep it filled. You don't really have to download the entire one-hour video in 15 seconds -- but there's tremendous value in being able to download
WOW; Way out west (Score:2)
Plan B (Score:2)
Use a carrier pigeon [reuters.com] :
Australia's problems with high-speed Internet can be summed up in one word: Margaret.
Margaret is a carrier pigeon that raced the nation's biggest broadband service to send a 700 megabit file over a distance of 132 km (82 miles) -- a televised contest that Margaret, with a memory stick taped to her leg, won easily.
Re: (Score:2)
I think this could work..
or maybe a brick and morter store that rents the media to end users.. then they wouldnt need the internet at all... this would totally bust Netflix's blocks... hey, thats a good name... Blockbus
Why? (Score:2)
Why should ISPs invest in infrastructure outlay when they can just raise rates on "bandwidth hogs"?
Re: (Score:2)
Because traffic keeps growing, even if you remove the excessive "bandwidth hogs".
Re: (Score:2)
**Winter month gas bill. Internet $130, Gas Bill $115
Don't forget the MPAA! (Score:2)
The biggest problem with this is that streaming allows for content publishers to control distribution (more or less). Allowing an entire work to be stored on a hard drive just begs to be ripped and stored permanently.
I hate the MPAA as much as the next guy. Just saying that even if the bandwidth was there, no way that the powers that hold the keys would allow it.
Out of curiousity.... (Score:2)
Out of curiousity, who will pay for this increased bandwidth/pipe? It seems that laying fiber everywhere across the country is either going to take government subsidies or be cost prohibitive if you are outside of metropolitan areas.
The second question would be is whether or not that increased bandwidth is the most efficient way to stream video? Dish network and DirectTV seem to do a pretty good job with video, now. Wouldn't it make more sense to have an internet connection from a satellite provider wher
Re: (Score:2)
Satellite bandwidth is much more expensive than fiber if you're not receiving broadcast streams.
Re: (Score:2)
Satellite bandwidth is much more expensive than fiber if you're not receiving broadcast streams.
But the example given was to receive video.
Re: (Score:2)
Turn the argument around... how much is a 1.5M ADSL line really worth today. What is a 100M link worth to a household? What are the differences in fixed cost per subscriber to maintain the infrastructure?
The capital investment may be significant, but that is how the industry has evolved; you just need to amortize it over an appropriate time period and the economics make sense. DSL doesn't have a 5-year lifespan remaining, which is why ATT and Verizon are putting in FTTH.
how is this not obvious? (Score:2)
The Internet is limited only by the capacity of the transmission medium and the memory of the switching equipment. I really wish more people understood this. Need more capacity? Upgrade infrastructure. Problem solved.
In a market economy, we could choose our own ISP by who offered the best service and speed based on price. Unfortunately deregulation has created a duopoly.
Re: (Score:2)
"Upgrade infrastructure. Problem solved." just shows how little you understand. You are talking about infrastructure that was traditionally amortized over decades (telco switches, cable headends, and outside plant) now being amortized over a few years instead. Sure, you can do that, but don't expect to like the price; multi-gigabit routers are not cheap (and it has nothing to do with the memory).
Just think! (Score:2)
Imagine how much more bandwidth they could oversell! Just imagine the possibilities! It's a win-win!
I have a simple solution to world hunger problem. (Score:2)
OMG those genius evangelists! (Score:2)
why aren't more people thinking along these lines?
Fundamental truths... (Score:2)
1. Data will always expand to fill drives no matter what size.
That data will be 90 percent porn. Higher capacities mean higher def. - ASCII in the old days to 3D High Def Stereo today.
2. Throughput will always expand to saturate the bus bandwidth.
That data will be 90 percent porn also. Higher capacities always meant higher def. From ASCII to gif and flv to 3D High Def Stereo.
Butts expand to fit the chairs they are in. This is what you get when watching porn all day.
--
BMO
Bandwidth fixes don't fix latency problems (Score:2)
Well, the internet probably does need more bandwidth to support Netflix [cringely.com].
And I'm not a fan of QoS to get better streaming video either. But is Cerf giving up on fixing the problems with streaming (and any realtime internet work) that we know about, bufferbloat [bufferbloat.net]? I heard about that from Jim Gettys (thanks to a tweet from John Carmack). Here's a two-page intro in IEEE magazine [bufferbloat.net] or a (more interesting IMHO) PDF slide presentation with nice graphs [bufferbloat.net] and there is other advice and documents and code on that bufferbl [bufferbloat.net]
Re:But won't that bandwidth just get eaten up too? (Score:4, Insightful)
We will not get 2k video, let alone 4k, any time soon. Most people don't even need 1080p. (With my set size and seating position I'm right on the line myself.)
Bursting the video rather than streaming it leads to less overhead. Clearly it matters how much less. If you can do it with UDP and don't drop many packets then the difference could be substantial. However, AFAICT most video streaming services utilize HTTP so that they can be accelerated by commodity caching systems.
Re: (Score:2)
Most video streaming services offer HTTP so they can get through firewalls. We are moving to a "port 80" world.
Re: (Score:2)
Only until DPI on port 80 becomes common and firewalls start blocking tunneling over HTTP.
Oh wait ... too late.
Re: (Score:2)
With TCP every packet is acknowledged. With UDP only missing packets are acknowledged and even then only manually. Any additional overhead is your fault. I agree with the notion that you would not do this except in dire need. Games still use UDP sometimes for this reason. Of course, if you botch congestion control then you're going to create MORE overhead, but again, this is your responsibility.
The only real reason to go to all the bother is if you are using multicast (that word is not even in the firefox d
Re: (Score:2)
I think what you are missing is the timeliness of the data. If you stream then any temporary slowdown, pauses or retransmissions due to packet loss, have a detrimental effect on the viewing/listening. Bulk downloads do not suffer this. Also, it can help even out bandwidth utilisation as you do not have the 'problem' of some periods when lots of people are streaming and other periods when the 'pipes' are comparatively empty.
Re: (Score:2)
I think what you are missing is the timeliness of the data. If you stream then any temporary slowdown, pauses or retransmissions due to packet loss, have a detrimental effect on the viewing/listening. Bulk downloads do not suffer this. Also, it can help even out bandwidth utilisation as you do not have the 'problem' of some periods when lots of people are streaming and other periods when the 'pipes' are comparatively empty.
What they need is something like disk striping on RAID drives, but for streaming video. If there were built in checksums embedded in the stream than maybe dropped packets could reconstructed without having to retransmit. Of course that would require a different video codec and make the data file larger, but if the goal is to stream video and the problem is lagtime, then it could actually improve the situation. Just a thought.
Re: (Score:2)
Something like that already exists:
http://en.wikipedia.org/wiki/MPEG_transport_stream [wikipedia.org]
No different video codec is required, just a stream protocol around an existing one.
Once 4K cameras become affordable (Score:4, Informative)
But people will just want to stream 1080p and then 2K
There isn't much difference here: 2K is 2048x1080, which is less than 7% bigger than standard 1080p.
and then 4K video
We can solve that once 4K video cameras become affordable for home use.
And can somebody explain to me why downloading a video in 1 gigantic burst is better than streaming it at a more steady rate?
If the entire work is cached locally, fast forward and rewind don't require a round trip to a server, and they don't require transcoding to create a new keyframe at the seek point. Nor will re-watching a video require sending it again.
Re: (Score:2)
if the internet connection is so fast that it can download the entire stream in 15 seconds, skipping foward/backwards won't be a problem
There is a limit to the speed of an Internet connection, namely the speed of light in copper or fiber. If your video server is on the other side of the planet, there's still a ping's worth of round trip to the server. Besides, accurate seeking requires a reencode of all the video between two keyframes in order to generate a keyframe at the seek point. This reencoding takes CPU time on the server side and reduces image quality by a generation.
nor will downloading it again
When your daughter wants to watch "Cinderella" three times in a mo
Watching the movie while away from the Internet (Score:2)
regardless of whether it streams or downloads in one chunk, it's still going to take me 100 minutes to watch the average film
For one thing, you don't have to spend those 100 minutes in front of a continuous Internet connection. If you can download a 100 minute movie in five minutes, you can choose the movie, download it, hop on the bus/plane, and watch it. I don't see the fact that home Internet is far cheaper than cellular Internet or especially in-flight Internet changing any time in the mid-term future.
For another, watching a 100 minute movie doesn't necessarily take 100 minutes. A video player supporting time-stretching wi
Re: (Score:2)
can somebody explain to me why downloading a video in 1 gigantic burst is better than streaming it at a more steady rate?
It isn't extra overhead in the protocol, its extra-overhead in the network equipment. Hardware that doesn't care so much about maintaining smoothly consistent transfer rates is a lot cheaper than hardware which does.
It's the basic design philosophy of the internet - stupid network with smart end-points. Implementing a smart network has all kinds of downsides. You can think of it* as being an NP-hard problem to solve if you try to do with a smart network, but NP-complete if you do it with a dumb network a
Re: (Score:2)
More cowbell!
Re: (Score:2)
Re: (Score:2)