Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway.
I always thought that was an Tanenbaum quote? The Wikipedia[1][2] citations use the word tapes but I seem to remember that DLTs was used instead of tapes.
This is why I labeled the quote "(apocraphal)". :) I first remember hearing it during the early days of Netflix---and I think it was Reed Hastings who said something similar to that effect as a quote of someone I can't remember now. I went with http://www.bpfh.net/sysadmin/never-underestimate-bandwidth.h... as the reference for Dennis Ritchie's variant of the quote.
This whole article is based on the premise of bandwidth scarcity. I don't buy it.
I also don't buy the complaint that suburbs are already suffering from too much interference. I live in a dense urban environment, and while I can see a lot of networks, I can easily get enough wireless bandwidth to max out my upstream connectivity.
Yes, radio spectrum is a scarce resource. But we keep getting better and better at using it. While there's an upper bound on the amount of data that can be sent over a shared medium, you can redefine the medium. Imagine if your wireless router only broadcast a low-rate unidirectional rendezvous signal, and then targeted each associated device with a focused high rate beam. Etc.
Not everywhere in the world has reliable landline connections and other forms of fixed infrastructure. And there are environments where wireless communications can be sporadically impaired due to weather or terrain (or ocean or space). There are also security and latency issues which might cause a user to prefer asynchronous communication.
There are fundamental limits on the capacity of a radio channel [1], but I understand we're not really near them yet. I haven't read the linked paper in detail, but I have attended a seminar by the lead author, where he gave a summary in plain English. I gather the limit is related to the surface area of the volume that the antenna occupies, measured in wavelengths.
I think it's a fascinating result, as in my mind such results are closely linked with some very fundamental physics and maybe even an ultimate information theoretic description of the universe. I'm probably overstepping the mark with that unsupported comment, but it is nice to think about.
This only presumes that you have line-of-sight or uncanny luck to have a reliable Fresnel zone to bounce signal through---but this is still constrained by distance. I used a "Portland, ME to Portland, OR" example to emphasize that these links aren't the kind that you can use allocated low-power radio spectrum to connect as in WiFi. You may be able to max out your upstream capability with local radio, but you can't supply bits to the other side of the city, state, or country. Unless everyone has the same exact interests as you, you won't even have the bits they want---this is what the pub-sub mechanism has to address on top of the content-network. Some people will specialize in carrying the bits that everyone wants and at every step of the way, nodes will have to look at subscription demand and throughput to decided which bits are sent in a batch. Yes, we can imagine our wireless routers as a low-rate signal system with focused high-rate beams, etc, but the probability that you will reach an interested audience from your radio is low. This is why FM/AM radio licenses are so valuable, and this is why having a content channel carried by all the regional cable monopolies is so valuable and vulnerable to censorship---there are many people who would love to watch Current.tv or Al Jazeera but cannot get it in their own regional cable systems. However, radio and cable are still synchronous mediums, which I'm suggesting we should back away from.
Using radio always assumes that there are other nodes interested in what you're sending. In nearly every case of radio-based Internet Service Providers I've been able to find, users always find that the service is "great" for the first month and thereafter totally sucks in every imaginable way: slow speeds, down-time, and bandwidth caps. There's only one tower and enough bandwidth for so many users, after that, there is no upgrade path because almost all the spectrum is allocated. This is guaranteed to happen for any homebrew WiFi network or municipal WiFi network---remember that you only get three separate frequencies without overlap ( http://www.bridgingthelayers.org/channel_overlap.html ) The limits are obvious with FM and AM radio is regulated by the FCC to prevent channel interference between cities---you never get "new" stations, you only get "different" stations when a broadcaster changes format. My favorite example of a broadcaster changing "format" is WHFS in Baltimore, MD---one day it was a progressive rock-radio station and the next morning a spanish-pop-radio station, without warning. (The DJs working there had no idea this change was coming.) By redefining the network medium to "bits via matter" instead of "bits via energy" (SD cards in my example), and relaxing our attention-spans to something less than "instant", you get more bandwidth by adding more cards to the network as needed. We may be getting better at using radio spectrum, but it will always have a political limitation of how it is used when it has real reach to many people---in short, if you can touch many people via a medium, more than a few people will always seek to control that access for their own interests. Bandwidth for information transmission via energy I still believe is scarce and will always be scarce for any situation where you wish to send to many people over long distances. The medium has to be redefined away from energy to matter---and perhaps, just to throw a huge monkey wrench into the mix and stir the pot more---this is what newspapers (which in their print format are already sending matter instead of energy to your household) need to add to their offerings to save themselves. Now, imagine receiving a 4GB SD card with your morning newspaper every morning while the paper-person retrieves old cards during delivery---and what would happen if you were uploading back to everyone else on those same cards? 8GB/day in and out becomes 56GB/week becomes 2.8T/year of data transfer, one SD card at a time. Relaxing our need for synchrony could make newspaper profitable again.
Subscribe to two newspapers and you double your network bandwidth and no one needed to climb on a roof. :)
Large data transfers are still best done through physical shipping of encoded atoms. Take a look at AWS Import/Export for one real-world example. But the threshold for that being useful is rather high. Have you worked out the breakpoint for where a person moving cards is more effective than a network connection? What information, other than DVDs, would be appropriate for this?
Why do you focus on radio-only connections? The highest bandwidth devices can easily be plugged into a wired connection. Why do you assume that bandwidth will get more congested? I think instead that there will be a move towards smaller cells, with lower power. The network provider could have an antenna on every other streetlight or power pole, where each only handles a few dozen devices in range.
Your comment about suburb wifi congestion doesn't feel right. The most cluttered wifi I've seen has been in big apartment buildings, where the transmitters are much closer to each other. One possible solution there is to move towards per-building based cells rather than per-apartment ones. Though this would mean that the hardware has to be as reliable as the building wiring, which isn't yet true.
Where I've heard uucp being useful is for things like "drive-by wifi" for rural parts of the world, for example http://www.firstmilesolutions.com/ .
But the idea that "relaxing our need for synchrony could make newspaper profitable again" is just strange. If people want their news less frequently then that just gives more time for their local network aggregator to download that news. Where's the cost savings or revenue boost for the newspaper?
I focus on non-radio solutions because I'm looking to see how we can do it without radio, period. So, there is no "live" connection for aggregators to download from. News is perhaps a bad name for the kind of content, think about it more in the sense of magazines or programs that aren't strictly time-sensitive, but entertaining and informative.
This is kind of related to the "Slow Web" movement I just read about in http://blog.jackcheng.com/post/25160553986/the-slow-web . I'm not interested in seeing regional radio-ISPs blanket cities with smaller radio. I'm coming from this with the view point that some people won't have an "active" Internet connection, period. This can be very rural or situations where the density is just too low to bother with radios---who's going to build a network for 5 customers? The networks create their own scarcity of bandwidth because they're only profitable where there are enough people---where there are enough people, there isn't enough bandwidth to give everyone a good experience. A asynchronous hypernet is something that is "instant-on" for any location where people desire to have it---there's no permitting required, building agreements to negotiate, pole rights to acquire, or hardware to buy-install-maintain. If you want it, a few SD cards, a USB SD card reader and software similar to UUCP to manage it will make it happen for everyone---the cost to start is less than $50 or even $25 for anyone wanting to start a node.
Instead of being potentially online 24/7, I'm imagining these asynchronous hypernet users are /never/ online with the Internet. What they receive is what comes on the memory cards or other mediums. If some content does come from the Internet, it's because someone elsewhere on the asynchronous hypernet is providing a gateway for it. For a news paper that isn't increasing their readership of physical papers and struggling to convince readers to pay for the online version (again, only useful to users of synchronous networks), a daily memory card of batched data will be attractive in any area where the local Internet monopoly starts charging significant money for GB of data above their "approved" threshold. If the newspaper can deliver data cheaper to the user than the synchronous Internet provider can, people will use it. If the newspaper makes their own original audio/visual content (which is already happening at newspapers like the NYTimes) and makes it available only via the cards, they have content that people can't get anywhere else---and a reason to subscribe.
That "drive-by wifi" shows one way to get internet service to rural parts of Africa and Asia, without other internet access. One proposal I heard had a guy in a motorcycle going from place to place with a wifi connection to do uucp transfers once in range of each destination. The other related technology you should consider is "delay-tolerant networking."
We can get wired electricity to just about all homes in the US, and it looks like there's telephone service available to 98%. Adding wired internet isn't that much harder than those other two, and there is a Federal program to promote rural internet service. Who's going to build a network for 5 customers? Us, through our taxes.
Iridium and other systems offers global service, it's just very expensive. I find numbers like $5/MB.
So you're talking about at most a small number of American homes which won't be able to have basic internet access if desired.
Of those, how many are interested in New York Time's data delivery service over newspaper delivery? I can't come up with any number which makes it interesting to the NYT - nor any other newspaper - to provide this service. Your point about caps doesn't affect things because the NYT iPad app takes between 1 and 20 MB per paper, depending on the number of sections read. Even daily reading of the entire paper is only about 0.5 GB and an order of magnitude or two less than the caps I've heard about. Print media just doesn't take much bandwidth.
Even worse, your numbers only possibly might make sense for newspapers with a reach into large amount of rural areas. That might work for parts of Alaska, but it won't help, say, the Burlington Free Press in Vermont. 95% of Vermont people have broadband access at home, leaving very few left who 1) want it, 2) can't get the print version, and 3) can't get the broadband version.
You say "where there are enough people, there isn't enough bandwidth to give everyone a good experience." In my experience this is rare, even in apartments where I've seen a dozen or more access points. ef4 confirms that observation. The link you gave also states that there is overlap but "This artifact has relatively little impact on the majority of 802.11 functions." Can you point to real-world cases where people are having problems because of too many access points? Can you estimate how often that occurs?
Overall, I think you're having fun with ideas. That's great. But you're letting your vision of what you want - you want people enjoying the slow web, perhaps - take precedence over what people really do want, and trying to back-fit justifications without a back-of-the-envelope estimation to see if those justifications are approximately right.
Suppose I live in rural Wyoming with no internet access, but wanted it. The first question is why? If I'm a farmer who wants to get better market information, then a slow internet won't help - the information might come too late. Okay, you can fix that - have someone come by every day. Is that the postal service? Will postal service holidays and interruptions cause problems to a farmer?
Perhaps I'm a writer, who needs occasional literature lookups. Perhaps then I could use the service well (assuming interlibrary loan and similar services aren't enough). But who's going to set up that sort of service, with the research staff on the end to find a literature reference?
But as described, your proposals are wildly off the mark, because people don't work that way.
If you want Internet access for news, research, and other "real time" information in rual Wyoming, of course you don't want the Slow Web experience, so this idea doesn't even apply. If the Slow Web is brought to you via postal service, you don't expect it everyday and its not a hardship if service holidays and interruptions occur. Do you complain for 29 days a month that your favorite magazine didn't arrive in your mailbox? No, you just wait for it, and enjoy it when it arrives that month. This is where most of the conflict I'm seeing with proposals for different kinds of radio-management are coming from. I know people don't "work" this way, but this was never an idea for the Fast Web to begin with.
The reason why I'm suggesting this is useful for newspapers isn't because this is how people would receive the newspaper, and only the newspaper, they would be sending and receiving any other content from the network (as in not-original-content by the newspaper) but instead from everyone else connected to the network via the newspaper. Newspapers are already suffering from the Fast Web/Slow Web problem, a newspaper on paper is slow-web and the newspaper online is Fast Web, and with so much freely available Fast Web content, they have a hard time convincing people to pay. News goes stale faster than the speed of rumor, so let the Fast Web have it, and make long-lived content that people want instead, enjoying it on their own time.
If access points no longer have the issues I observed 4–5 years ago, that's great, but they don't solve the problem of network coverage over larger geographical areas and I'm pretty sure that few or no access point systems can handle the load if someone fires up a popular torrent or watches a streaming movie on the shared network, but then, no one wants them to do that, that's a violation of the Fast Web Etiquette—but receiving an hour television show you subscribe to on a 4GB SD card is going to be a much better experience in terms of video quality and the lack of buffering should there be any hiccups with the Fast Web. In that respect, newspapers can become the "store" for other desired content and save their collective asses.
My point was to ask you describe the sorts of information which are best sent via the slow web. I gave some examples where it wouldn't work, because of the need for timely responses.
Your example of receiving an hour television show on a 4GB card does not work, because we already have that. It's called Netflix. I mean, it's a perfect example of what you're wanting, but it's already a solved problem built on top of the postal system, and not based on uucp nor your proposed slow web. There's nothing stopping you right now for implementing this via the postal service - so my question to you is, which types of people would want this sort of system, why would they want it, why will there be a long-term demand for this sort of service, and why doesn't it already exist atop the postal system?
It looks like you propose that the newspapers could leverage their delivery system in order to deliver the data. But surely the postal system is the better fit, given that newspapers don't deliver to place with "only five people" and instead charge more so they can mail the newspaper.
The other problem I have is that you keep coming back to "radio-management." That's not important. We can wire 99+% of US homes with wired connection - we did that with power and are nearly there with phones. Why is radio important if you can get 14.4K wired to your home? Assuming a rather low value of 1KB/sec over twisted pair, that's 2GB per month. If you also want local wireless for the house, that's under $100. I set up that system 10 years ago for myself.
To emphasize, 1KB/sec is fast enough to download the entire NYT in 6 hours.
So the potential users, after removing the Netflix factor, are 1) those people who can't get any network connection, and 2) those people who are restricted to dialup but want more than 2GB/month, and don't care about 'fast web' responsive access. Both are small and decreasing in numbers.
Oh, and how does someone signal which 4GB of data they are interested in? I tried but completely failed to come up with any solution which doesn't require a large majority of the sites in the world to rearchitect their interface for this tiny minority - and that's not going to happen.
To be short: Netflix doesn't accept uploads from customers via postal mail—so it doesn't exactly match the example. Most people want content that requires more than 2GB/month. Yes, NYTimes is much less than 2GB/month, but this wouldn't be about receiving "just" the NYTimes---the newspaper delivery service would work as a "carrier" of bits. Users signal for the data they're interested in via what they upload. Publishers of bits upload indexes to the content network. It can be implemented now with the postal service, by sneakernet in communities of people, or it could be something that newspapers implement, it doesn't matter who does the physical work of bit-moving. And finally, I'm not interested in supporting the "large majority of the sites" because they're all Fast Web. If they want Slow Web, they can add it, but they don't have to, it's not mandatory and its not expected of them.
My examples were all meant to show areas where a slow web would not be effective. Who are the people and what are the use cases for a slow web, and why doesn't it exist now on top of the existing postal system?
The newspaper delivery service would not work. Period. Delivery service only covers the local service area, where there is a high enough density of subscribers to make a special delivery service worthwhile. These people can get broadband and/or cell service. Otherwise newspapers use the postal service, and charge a premium to those customers who want that service.
Now you say that "upload" is important. Yes, when my wife was deployed in Iraq, we mailed USB sticks back and forth, containing music and video files. So that service exists now for point-to-point, and scales decently well to several people. Who are the people who want one-to-many data broadcasting via the slow web? What do they want to accomplish?
And I return again to mechanics. How does one "signal for the data they're interested in via what they upload?" If I send videos of our baby to family, do I get baby videos in return? Or do I get ads for baby products? I can't think of any mechanism which would handle what you are describing, other than having a human concierge service - and that's expensive.
If there are no adapters to make the slow data resources which are available on the fast web available to the slow web, then why would people use the slow web?
Be part of the slow network movement? I was so excited when a friend agreed to connect to me and bring in email and usenet via UUCP. Good memories, but I like the real-time fast food world a lot better. Though for interplanetary communication it makes a lot of sense.
Actually, Usenet worked amazingly well even under the limitations of computers and networks in the 80's. Shared information streams strike me as superior to isolated data motes and walled gardens.
It used to be the case that device nodes for serial ports and lock files for those were owned by that user/group. Historically uucp (uucico - uucp call in call out) was the only sensible thing to use the modem for anyway...
Personally, I never used uucp over serial links, but for several years fetched my emails via uucp-in-tcp. It was a homegrown system with sendmail, later qmail, later exim and it worked very reliably.
Encryption would of course be mandatory. Maybe even enhanced with steganograhy.
"My dear photo sharing friend, enclosed you'll find the collection of my latest pictures which are simply to huge in volume for my crappy internet connection. Enjoy and keep sending pictures in return. PS: I love your artistic subtle casual picture style."
I always thought that was an Tanenbaum quote? The Wikipedia[1][2] citations use the word tapes but I seem to remember that DLTs was used instead of tapes.
[1] http://en.wikipedia.org/wiki/Sneakernet
[2] http://en.wikiquote.org/wiki/Andrew_S._Tanenbaum