Almost ten years ago, I read a fascinating essay by George Gilder in Computer Underground Digest, reprinted from Forbes. He predicted that in The Future, technologies like the erbium doped fiber amplifier would cause bandwidth to grow faster than CPU power, and that that would eventually have drastic effects on how we do computing and networking. I'm not sure that prediction has really come true, but supposing it did, what would be the effect on the copyright wars?
The thing is, one of the most infamous forms of DRM is the artificial distinction between "streaming" and "downloading". In order for you to watch or listen to audiovisual content from the Net, the bits representing that content must be transferred to your computer. The technical term for that is "downloading". It's then up to your software to determine whether to hang onto those bits after you finish watching or listening, so you can watch or listen again in the future; or delete the bits after they've been downloaded and used once. Deleting the bits as soon as they're used seems to be what people call "streaming". They've even managed to turn "download" into a dirty word for some legislators. But the download must happen first, it's part of what "streaming" is, and the difference is entirely on your end.
The media cartels want to charge you more for "downloading" (which they define as NOT discarding the data after a single use) and "streaming"; and they want to enforce that on their end, at the server, by somehow making it impossible for me to NOT delete the bits after looking at them once, if I don't pay extra for the privilege of using my own computer. These attempts make me angry because they charge more for a service that costs them less work - sending me the bits once is the same act on their part no matter what I do afterward, and attempting to force me to forget actually costs them a lot of work - and they're denying nature to my detriment and their benefit with no justification except a series of appeals to emotion and other rhetorical fallacies.
Their DRM fails in preventing me from remembering the content after I've downloaded it the first time, because bits have no colour. There's nothing they can do to the bits that will make them evaporate after one use. I can always just choose to ignore the instructions they send that purportedly cause my computer to create the phony distinction between "downloading" and "streaming". But that analysis, which leads me to conclude that DRM is necessarily doomed to fail (and so attempts at DRM are insults to my intelligence) depends on the assumption that I have access to effectively unlimited amounts of computer power. It works only because modern computers are cheap and plentiful.
One of the media cartel's responses is to try to prevent me from having access to computer power in general, by attempting to pass laws that would force all computers to be crippled and unusable. That's the basis of "Trusted Computing". Unfortunately for the media cartel, computers are just too useful, and too easy to build; there's no way a law to keep them out of the hands of people like me, could ever succeed. But what if instead of attempting to create a human law, they fell back on the laws of physics to do the work for them? Those laws are a lot more solidly enforced than any human law can be.
Here's the thing. At the moment, network bandwidth is pretty expensive in comparison to the sizes of the things we want to send over networks. A Hollywood movie, in the most natural digital form of what actually would appear on your computer screen, is maybe two hours times 24 frames per second times 1280x1024 pixels per frame times three colour channels times eight bits per channel. That's 5.43 trillion bits. The network connection into your house, if you've got a good one, might be as much as 4 million bits per second. Downloading or "streaming" the movie in its most natural digital form would max out your connection for 1.36 million seconds, or about 16 days. You can't do that if you want to watch it in real time, and you won't want to do that if you can walk down to the video store and get a disc in a couple hours. The movie is also 679 billion bytes (which marketers call "gigabytes", though a genuine gigabyte is a little bigger than a billion bytes), and the price of hard drive space to store that and play it back in real time, isn't trivial.
Everybody who wants to send movies over the Net faces the bandwidth wall, whether they're the studios or the pirates or the fair-use guys. So what they do is they use mathematics to compress the movie down to a much smaller package, maybe only a few hundred megabytes, which can be shipped over the Net in reasonable time. Compressing it on the sender's end takes a lot of CPU power, but they only have to do that once and store the result; decompressing it on the receiver's end takes a fair bit of CPU power, but current PCs are overpowered anyway and so the bandwidth, not the CPU, is the limiting factor. And this process is what supports my "DRM is impossible" claim, because any file small enough to fit over the Net is by definition small enough to fit on a hard drive.
But what if Gilder is right, and bandwidth is growing faster than CPU and storage? There's a lot of reason to think that that is true or eventually will be true. It's a fact that in 1997 or even in 2007, we're not really at the limits of the physics for communication speed. Lots of fiber is still "unlit", that is, not being used for communication. Even the "lit" fiber seldom carries anywhere near as many distinct "lambdas" (laser frequency channels) as it physically could. And hardly anyone even has fiber for their "last mile" at all, and that's more because of the failure of the free market than any technical limitations; the phone company is happy to stick with extremely expensive technologies as long as they can make a good profit, instead of switching to the latest and greatest and being forced to lower their rates accordingly. So it seems like there's lots of room for growth in bandwidth.
There is much less room for growth in CPU power and storage capacity. On those fronts we're running into really hard technical challenges, not just "We could run more lambdas but don't think it would be profitable". The CPU on my desktop is a few years old and runs at 2.66GHz. I think the latest and greatest is like 5.1GHz. Maybe if you're an insane overclocker using the latest developmental stuff you could run a similar CPU as high as 10.0GHz for a few seconds. But you can't buy a similar CPU to run at 100GHz at any price; they just don't exist and possibly never will. Similarly, on the storage front we used to use hard drives based on "magnetoresistence" and then it looked like we had pushed that as far as it would go until someone discovered "giant magnetoresistence" which allowed, I don't know, say another factor of ten in storage capacity, and then that hit its limits and they discovered and started using "collosal magnetoresistence" which allows maybe another factor of 5... and unless there's another revolutionary scientific discovery, we may actually not be able to get much more out of magnetic storage media. Even if there is such a discovery it probably won't be as big as the previous ones, and it won't happen as fast as the increase in communications bandwidth.
So at the moment CPU power and storage space are big in comparison to communications bandwidth, but it seems like a serious possibility that eventually, that balance will be reversed. What happens to the downloading and streaming situation when the balance is reversed?
Well, it won't make sense to compress media so much. Instead of shipping you the compressed version of the movie, the studio could ship you the raw bits, all 5.43 trillion of them. To do that in real time as you watch the movie would require 754 million bits per second of bandwidth. A fiber can carry billions, easily, with current technology, well within the physics, so it's only a question of rolling out those networks and doing it at a price you can afford. Having 754Mbps of bandwidth to your home computer would be prohibitively expensive at the moment, but it's a meaningful thing to ask for; the answer is just "you can't afford it", not "it doesn't exist". In the future, maybe you'll be able to afford it.
What to do with those bits when you get them? The computer you have today wouldn't be able to pull that many bits off the network and put them on your screen, but it's not far out of reach; within a few years that should be do-able. Remember, it doesn't have to spend time decompressing, because we're assuming that the bits are sent uncompressed (and unencrypted, by the way).
But putting the bits on the screen will be all your computer can do. You won't have disk space to save them for later viewing. You won't have the CPU power to compress it to fit on disk. Even if you buy a bigger CPU and disk, by the time you do that you'll also be buying a faster Net connection, and the studio can make the movie bigger. With bandwidth growing faster than disks and CPUs, they can always eventually win that race, and send you a movie too big to store, too big to compress, fundamentally, too big to pirate. So at that time, they will be able to enforce a "view this once but do not save it for later viewing" restriction by eliminating all their foolish (and CPU-expensive) compression- and encryption-based DRM, and just feeding you so many bits that saving them is impractical.
That's the same way that, until recently, live concerts "protected" themselves from "piracy" - the experience of going to a concert required so much bandwidth that an amateur just couldn't carry in the equipment needed to capture it and really do it justice. Even if you made a "bootleg" it was nowhere near being there, and concert promoters who weren't batshit insane didn't spend time worrying about "bootlegs"; they served as advertisements for the actual concerts. The concert-taping situation changed because consumer recording equipment became smaller, better, and cheaper; that's why you now see posters up in movie theatres full of lies about the evils of camcorder "piracy", which wasn't even a possible thing to attempt until quite recently. But the balance for network-delivered media may be shifting the other direction. With bandwidth cheap in comparison to CPU and storage, "piracy" of "streamed" content could fill a similar niche to old-school bootlegs - a reduced-quality advertisement for the full "streaming" experience, because it's just not practical to copy and share the "stream" at full quality. And just because of the laws of physics, with no artificial DRM involved.
Of course, that's all conditional on bandwidth really becoming cheap relative to CPU and disk, network neutrality remaining in existence (because differential pricing would kill high-bandwidth streams before they started), and the movie studios seeing the situation in the same terms I've described, and setting their prices low enough to really make streaming look good in comparison to saving for later viewing. I'm not holding my breath for those things to happen. But it's one vision of a possible future I wanted to share with you. Please be aware that I'm not saying the world of unlimited bandwidth driving out local storage is a good or bad world, or one that we should be trying to achieve or trying to avoid. My point is that if these trends are genuine and continue, it's the world we will all end up in - everybody, the studios and the pirates and the fair-use guys - whether we like it or not.
Comments
owen from 74.120.31.7 at Wed, 17 Jan 2007 05:20:11 +0000:
But maybe we'll have like, meat computers in our heads that can store the data holographically on like quantum grains of data sand or something that compute in different parallel universes at the same time and all the movies are, like, already existent in our own brains and the hardware just like brings them forward into the concious mind, right?
Bob from 216.39.176.80 at Wed, 17 Jan 2007 13:18:12 +0000:
Wow, owen, you saw the Matrix too? :-)
Chris H-C from 64.235.97.125 at Wed, 17 Jan 2007 15:23:32 +0000:
That's not really the issue of this article, guys. Matt's talking about the *trends* of growth. If you have a math background, think limits.
If bandwidth gets faster at a higher rate of speed than processors and disks (no matter how you judge speed or capacity), then eventually there will be something someone wants to sell you that will fit on the ultra-fat pipe but won't fit on your drive, holographic sand particles, or what-have-you at full rez (the disk part). And since there is just so much raw data coming down the ultra-fat, the processor or quantum meat computer just can't encode it down fast enough.
Now, this may not start to happen. While all this is going along, mathematics will continue to get better and new encoding methodologies might crop up... Actually, just about anything *might* happen. Matt was just discussing what *would* happen given his assumptions.
It's a little like playing Imagination, but only one imagination's allowed.
Steve from 165.154.153.142 at Wed, 17 Jan 2007 22:43:34 +0000:
I disagree... with Forbes, not Matt. We are close to the end of transistors age, and starting the nanotech age. In the same way that leaving behind vacuum tubes caused a computer/communications revolution, leaving behind transistors will do the same thing.
A guy in the Prairies has already created single molecule transistors. (I forget if I saw the story break 1 or 2 years ago now.) But regardless, telephone companies have no desire to put in that last mile of fiber, and still won't long after nanotech becomes common.
Matt is absolutely right, it is a failing of the free market that we don't have better telecommunication pipes right now. More importantly in terms of the future... there is no reason to expect that the free market situation will change and telephone companies will *want* to lay fiber. So the status quo of crappy wires still being the limiting factor will continue.
Ironically, today I was arguing with a Bell tech who came to my house who said that if he had found the problem, there would be no charge. But because he wasn't able to find the problem, then there would be a $100 charge. Yay... I get to pay for Bell's incompetence... again. (My street can't even have DSL... so fiber is going to be 30+ years.)
Vilhelm Sjöberg from 217.146.104.65 at Thu, 18 Jan 2007 21:34:00 +0000:
Interesting idea. In effect we would end up where we were with television in the 70s: lots of bandwidth in the luminiferous aether, very little intelligence in the recievers, and no storage space on the receiving end.
A few random thoughs:
(1)
As Owen says, computing power has probably not hit a hard limit just yet. Even if clocks never reach 100GHz, video compression should be quite amenable to parallelization, certainly up to at least hundreds of parallel tasks, and computers using hundreds of 2GHz processors are quite feasible. Indeed, you can already buy them off the shelf.
(2)
Even if bandwidth does increase faster than than computation speed and disk space, another crucial assumption to get this "natural DRM" effect is that the effort to do compression will be dramatically bigger than to do decompression. After all, we assume that the computers doing the broadcasting have space to store the content, so hardrives can't be all that limited. For mp3s it only takes perhaps 10x more effort to compress than to decompress, and the march of progress has ensured that computers can do *either* in real-time without much trouble.
To be sure, you can invest arbitrarily much effort in compressing things, but the returns diminish. If the content you are compressing contains stochastic effects (like thermal noise in the sensors of hypothetical very high resolution video cameras), it may turn out that quite simple statistical coding schemes already come close to the Komolgorov complexity.
Loss-less compression of photos is apparently like that: the least significant bitplane in a "truecolor" (24-bit) photo is almost indistinguishable from white noise, so no matter how much intelligence you put into scene analysis etc, you don't get a compression factor better than 8. So this puts a limit on the coding/decoding asymmetry: even if the hypothetical photo pirate stored the images completely uncompressed he would not use more than 8 times the storage space that the broadcaster used.
Lossy compression admittedly changes the game a bit, but the general principle seems to hold: if the data was small enough that the broadcaster could store it and copy it in a convenient form, there seems to be a fair chance that the consumer could do that also.
Of course, in a high-bandwidth world, one can easily think of applications where it would be unfeasible to store the resulting datastream at all. For instance, "Google Earth, realtime edition", where we have a very large number of satelites all constantly streaming high resolution video data. Anyone could at any time take a peek at any place, but saving all the data for looking at later would be daunting. But this is not very much like DRM: the laws of nature don't permit *anyone* to copy that data, not even its rightful owners.
(3)
How many bits are available to bury us in anyway? Some content doesn't use very much bandwidth to begin with.
The sheet music that was the focus of copyright wars in the 1880s perhaps come down to 10 kilobytes for a song. By the 1980s, BBS systems could pirate the texts of novels (hundreds of kilobytes). In the late 1990s, computers could comfortably swap around pictures and song recordings (megabytes). And now, we have reached to point where we can handle movies (gigabytes, or perhaps hundreds of gigabytes in high-rez). Are there any new art forms around the corner to saturate our fibre-links?
And of course: people still use sheet music, the phonogram didn't stop people reading books, and the invention of movies has not killed off music. It seems that the set of stuff that is technologically feasibly to copy and store grows to include a bigger and bigger portion of the stuff that is actually produced and enjoyed.
In other words: given how fast computer technology progresses, the limiting factor is no longer bandwidth or disk space, but the actual process of creating the bits in the first place. Unless we can make the supply of artists and film directors somehow follow More's law, computer systems will be quite able store anything that human creativity can produce.
(4)
Fast networks changes more than just the delivery mechanism. If bandwidth became much bigger than disk space, we would certainly not each want a personal copy of a movie (or a sim-stim stream) sitting on our own harddrive. Network filesystems would be a big hit, with all the computers on the planet pooling their available space.
In particular, if disk space were more scarce than bandwidth, it would be expensive compared to other computing resources. In that case, a business model based around storing lots of bits yourself and streaming them to people on-demand suddenly has a higher cost than one based around distributing the bits to as many different people as possible, and economic incentives drive your towards resurrecting Napster.
Just as today, the cost of storing the products would probably be essentially negligably compared to the cost of producing them in the first place (the cost to produce a physical CD being measured in pennies, and all that), so that incentive in itself would not create communism. But it does shift the balance in an interesting way: millions of consumers working together have a lot more disk space and cpu power (the only computing resources that count for anything in the future) than even the biggest record company.
If fast networks provide the means and the necessity for them to work together, why, imagine what could happen...
Anonymous from 200.157.204.10 at Thu, 03 May 2007 03:01:05 +0000:
Your idea cannot work. If you have an ASIC fast enough to receive the data and sent it to the display, you can also make an ASIC which is fast enough to decimate it to a low enough bitrate that another ASIC can compress enough for you to store. In the worst case, you can just drop 9/10 of the bits and save the rest, without bothering with the antialiasing.
Tim from 130.88.202.5 at Thu, 03 May 2007 09:29:43 +0000:
With lots of bandwidth, novel forms of clustering might take off. Some people have demonstrated ways to use ICMP echo and other protocol side-effects to generate a memory effect from a network. It's not much use now because storage is more plentiful than the bandwidth needed to make an equivalent.
BlueNight from 205.201.141.146 at Wed, 16 May 2007 21:01:36 +0000:
The history of civilization is the history of tools.
As the tools improve, the existing laws (built with a view of current tools only) adapt or fail.
The FCC can regulate the content of radio and TV because the airwaves are a "limited public resource". It doesn't count as federal regulation of speech or the press (breaking the First Amendment) because of these technicalities. At the time it was written, speech meant exactly that: speech. At the time it was written, the press (and hand-written letters) were the only tools for recording and transmitting one person's ideas, opinions, etc.
Intellectual property (IP) is on the opposite side. People who have the legally-recognized right to control the use of such property must rely on legal means to do so. Encoding media so that it can only be decoded by their players is currently one such way; non-duplicable media is another.
A bound book is hard to duplicate. However, it can be done by cutting off the spine and putting the pages through a scanner (with a page feeder). Only pirates are likely to do that, because the reprinting cost (without a printing press) is likely more than the value of the book. (Figure 9 cents a double-sided page at some copy places: a 350 page book would cost $15.)
The less easily duplicated the medium, the more easily pirates (for profit) are differentiated from those whose interest lies in the enjoyment of the medium.
When a tool is made for one purpose, and that purpose is illegal, possession of that tool is prima facie evidence of intent to break the law. When a tool has multiple uses, but all legal uses are rendered null by licensing, only illegal uses remain.
mendel from 87.139.35.219 at Mon, 09 Jul 2007 21:56:40 +0000:
The people who invented CD-ROMs thought that way: "there's much too much storage, there aren't any media that'll hold this much information except for tape drives, and they're too cumbersome". That kind of reasoning was good for a few years, then cheap CD burners arrived and that attitude went the way of the dodo.
If the studios have the hardware to hold your proposed gargantuan decompressed movies in their on-demand video libraries, how long do you think it will take for that hardware to be comsumerized?
Matthew Skala from 69.63.63.131 at Thu, 12 Jul 2007 02:19:07 +0000:
One thing they could do would be to store it compressed, then decompress it to stream it. Video compression is a lot more expensive than the corresponding decompression, so the fact that the studios can afford to decompress it on the fly wouldn't automatically mean that consumers could afford to compress it for storage on the receiving end. You're right that that kind of argument does limit the effect that very fast communication could ever have, though, because *some* stage in the process will almost certainly involve storage no matter what.
Add Comment
Do not enter a fake email address. If you don't want to provide one, just leave it blank. Comments with fake email addresses will be deleted.
This form is for posting public comments to be read by other people who visit this Web site. If you have a software support question, or other material directed to the page author instead of to the general public, please send email instead.
All the data you enter, and your IP address, will be saved and displayed. Don't enter secret information. HTML is not accepted; it will be displayed as plain text. Your comment will only be added if you enter valid data in all required fields; if it isn't, use the back button and try again.
I, and I alone, reserve the right to remove postings for any reason.
![[RSS syndication file]](/i/rss.png){kind=small}
Ghap from 69.158.162.211 at Tue, 16 Jan 2007 19:46:21 +0000:
I don't buy.
Some notes on GHz:
- You can get a much higher GHz rating, if your chip is very simple. The limiting factor is the speed of the signal going between transistors; if there were few transistors to travel through for each cycle, you can get more cycles.
- As an AMD advocate, I will remind you that AMD processors have had more computing power per MHz/GHz than Intel since well before they were using the term GHz.
- Moore's law in its correct form doesn't say anything about cycles per second. It says that the number of transistors will double every two years. This can still be true if the GHz of the chip is not increasing.
- Dual core systems certainly put out more computing power at the same GHz.
My point: GHz is actually irrelevant. Even if we reach a GHz cap, this doesn't mean computing power per CPU will decrease. If chip manufacturers hit a wall for power per CPU, they would switch focus to power per $. Future computers will likely have a large number of cheap CPU cores running concurrently, while costing less than current one (or two) core PC's.
I haven't crunched any numbers, but I suspect that before you would see huge amounts of cheap bandwidth like you talk about, you would be able to compress an HD signal real-time. I imagine current hardware can compress an HD signal real-time right now, if it first reduces the resolution. Now, I might believe that you wouldn't be able to compress the image at full quality, but modern formats really give more quality than is necessary for most purposes. This will likely be more true for future formats.
Interesting thought experiment, though.