Homes with Tails: Homeowners providing their own fiber

In an audacious new paper, "Homes With Tails," Tim Wu and Derek Slater argue that there's no good technical or economic reason why homeowners couldn't supply their own fiber-optic internet connections that run hundreds of times faster than today's connections:

We call this property model “Homes with Tails,” for the fiber would form part of the property right in the home. Key facets of our approach include:

1. A “condominium” model for fiber ownership, in which individual strands of fiber are sold to consumers, while maintenance and other collective needs are managed jointly.

2. Private firms and municipalities could consider selling fiber connections based on this model; and

3. Governments could consider using various mechanisms to support consumer purchases, including a tax credit to homeowners or renters who purchase a broadband connection.

Summary: Homes With Tails, PDF: Homes With Tails



  1. I’d really like to see this become a reality. With every big ISP looking out for #1, which in turn means screwing #2 to infinity — us customers, it just might make the internet pick up more speed in it’s development. Now if we could only get it done.

  2. Ww wht n dt. Attaching a fiber to your house doesn’t do a hell of a lot: Consider the innocuous-looking “trunk” labeled there. If the telcos don’t build fiber out to the pedastal, then it makes no difference if you’re reaching their copper with OC-768 or whatever.

    Moreover, even in most access architectures (all, basically)you don’t have a dedicated fiber all the way to your CO (Central Office), but hit some router or PON node or other aggregation point where you share the bandwidth on the other side anyway. But if your telco hasn’t built that out, then your little fiber hanging off the outside of your house is useless anyway.

    No, the last 100 feet are not the problem, but the last mile is, and you can forget about consumers paying that.

    Need we mention the issues and costs associated with powering the return signal’s laser? And then, of course, you have to have a house wired properly for this to matter.

    This is a niche application at best, though it might make sense for RICH condo villiages/gated communities to do something like this.

  3. My friend was in Ghana for 2 years in the peace corps. The house he wsa provided with by the local chief had lightswitches and fixtures, even a fuse panel….

    ..and was 25 miles from the nearest power pole.

  4. Uhh, check out how Japanese telephone lines (used to) work. You would buy your connection, and then pay for the monthly service. It ended up selling more of those advanced cell phones we often covet.

  5. Do some research and you’ll find:

    -The minute (inch?) you cross a property line with a cable you become a utility and are regulated by a state’s utility commission.

    -The incumbent communications providers will not provide service over anyone else’s infrastructure. (ie. build a development with first class infrastructure and Verizon will not put dialtone on it.) You need to buy services like any other business and manage your own infrastructure.

    These rules are the ones that need to change. Call your local utility commission and state representatives.

  6. Yeah, I don’t know how many homeowners are going to get past the sequel to “Fiber for Dummies,” the 1,100 page masterpiece, “Getting Sued by your RBOC for Dummies.”

  7. Keeper, RTFA. I mean, seriously, RTFA. What a stupid thing to write.

    Witczak: Tim Wu is one of the most respected telecoms scholars in America. I suspect that again, if you RTFA, you’ll find that this has not escaped his attention.

  8. Well, the *entire paper* is about why this isn’t the hard question that Keeper makes it out to be — you make not agree with them, but to say, “Wow what an idiot. Attaching a fiber to your house doesn’t do a hell of a lot” is to clearly indicate that you didn’t read the paper, particularly the section labelled “Interconnection, and Who Provides Services?”

  9. Broadband is a utility, plain and simple.

    It makes a lot of sense if you replace every wire into the home (with the exception of power) with a single drop. Fiber, coax, or copper, or even wireless. It carries your TV, your Internet, your phone… not that there’s a real distinction anymore.

    Point is, one regulated drop, with the physical connection leased from and managed by a company who’s only role is to make sure the data flows. They provide no services or content, all they do is connect. Much like a landline or your power, you choose the service providers, pay their rates, get their content.

  10. Cory:
    Well, I’ll check it.

    I used to work with Paul Shumate at Bellcore (Bell Communications Research) and I was for a number of years a fairly well-known expert on multiwavelength optical backbone transport (regular talks at NFOEC, a couple of invited papers, even founded an optical access company…
    Here’s one of my papers:

    So up front it didn’t even look worth reading, and I’m still not so sure.

    I’d point out that, even if wrong, the comment wasn’t “stupid”: The issues I raised are precisely the known issues in fiber access architectures and the writeup above made it appear that the authors weren’t even aware of that.

    Do you now want to debate the merits of burst-mode optical transceivers in PON access architectures, or would that be “stupid”…

  11. OK, I read the F-in article.

    Sorry, not hugely better in the reading. The guy doesn’t understand technology and it shows. In fact…

    “Tim Wu is a Professor of Law at Columbia Law School, a Fellow at the New America Foundation and also serves as Chairman of Free Press.”

    Don’t get me wrong, on the surface the arrangement would have some appeal, but it just doesn’t map to the economic or technological realities very well. Should I go into detail? No, but a couple of quick hits:

    1. The “10 Gigabit club”: He doesn’t seem to realize that OC-192 (never mind anything faster)is incompatible with aerial cable due to Polarization Mode Dispersion. Not exactly “stupid”, but a major newbie error for anyone in the fiber world. This alone blows his cost “argument”.

    2. (Here’s the big one) His statements about Moore’s law applied to access bandwidth (and pretty much the entire premise of the article) indicate that he doesn’t understand the basics of packetized communications and shared media. This is the reason why it may never be desirable (never mind cost effective) to have a dedicated fiber all the way from the POP to your house: It’s extremely rare for even the heaviest of apps to burst up to even a far smaller pipe’s full bandwidth, and if it happens (and there’s contention from other users), the upper packet layers just retransmit. There’s a name for this: TCP/IP. It’s also why the internet works.

    3. Bill St Arnaud.
    He claims that Bill St Arnaud has been championing this model for “over 10 years”…is this the same Bill St Arnaud that worked at Tyco Submarine Systems until the early 00s?

    Note relatively few references to actual known fiber journals or even access standards.

    No, it’s on the surface a cute idea but really betrays a complete lack of understand of not only access architectures but even the nature of modern packetized communications a la the Internet.

    My original comment was not far off.

    So, Cory, I “RTFA” but I judge it to have been largely a waste of time.

  12. Lantern:
    Yeah, but opening with “wow, what an idiot” probably didn’t help a fair reading of your criticism.

    I’m genuinely interested to hear what you make of the article though. Sounds like you might be someone to pull out the pertinent questions.

  13. I have idly dreamt of a “cobweb” of self organizing nodes (cobs?), where I run a fiber to each of the neighbors I can talk into connecting. They do the same, setting up a NaN. But then I start to think ahead to our NaN getting filled with spam and malware and trolls and such, and the idle dream sort of fades.

    But as NORES mentioned, your RBOC is going to damn it with faint praise at best. In the bad old pre-divestiture days the attitude of Ma Bell was that she owned the lines, including the wiring inside your home, and you merely rented the black rotary dial terminal from Western Electric. They have “not for sale” stamped on their bottoms, heh. What with the re-assembly of the Bell System, that notion is arising from its grave to wander the land again :)

    Anyway, cool idea. I wonder how much I can charge the local ISP to peer with my neighborhood area net?

  14. What’s so strange about it? Here in Sweden the system is exceptionally common, for bostadsrättsföreningar (basically condominiums) at least. The homeowners’ association gets a switch and a land-based fibreoptic line to some ISP, like any company would, and broadband is included in the monthly membership fee. Or like my current homeowners’ association, just get an ISP to set up and maintain the whole thing.

    How do corporations in America ever connect up with fibre is what #5 says is true?

  15. You do understand that laying underground cable cost $5/foot (fully allocated) back in 1980? It’s probably more now. So, say $10/foot in 2008. It would add significantly to the cost of development. I don’t think it “pencils”–if it adds substantially to the costs of buying a house only well-to-do people will buy it in preference to just having service installed by a major carrier.

  16. Further insight might be gained by looking at the Ruby Ranch Internet Coop. It has done something similar with the technology of the previous decade, employing locally owned copper to members’ homes. (One advantage RRIC had is that a bunch of its members are lawyers).

    And, #17 : wow only $10/foot? My neighbor is 10 feet away. Woohoo, free internets forever for only $100!

  17. #13: Well, I guess I agree. But then again, on the surface it appeared as if some lawyer or somebody had dreamt up some scheme without having done any homework on the engineering. After reading the article, it became apparent that a lawyer had dreamt up a scheme withbout having done any homework on the engineering.

    #16: Actually, what you describe is not strange or unreasonable, but it’s not really what the article was suggesting. The article focuses on running fiber all the way from the CO to the home using the “home with a tail” model. That’s entirely and completely different, and not only infeasible but (most likely) pointlessly and impractically costly.

    #15: There’s actually a (as far as I’m concerned) far more interesting model along the lines you are talking about. To date, all the WiFi home nodes operate in a star configuration. WiFi has another “lateral” mode that allows nodes to move traffic between them. In other words, WiFi allows for your home router to route packets directly through your neighbors’ router if the two WiFi clouds overlap, but the feature isn’t being used currently. Someone (I can’t remember who) pointed out that this could cause there to be a lot MORE bandwidth, even without anyone being hooked up to service providers, as packet traffic jumps from WiFi node to WiFi node.

  18. For those of you who have read this far, I’d like to just point out something which may be of general use.

    Keeper of the Lantern is arguing from a “these guys don’t know nothing about optical fiber” point of authority, which sounds pretty convincing. Except for a) I know Wu + Slater will have run this past other fiber experts, and b) anybody here who knows anything about Wi-Fi, which is probably more than know about optical fiber, will recognise that KotL’s last comment about a special “lateral” mode, is an incredibly confused explanation of ad-hoc mode and mesh networking, depicting that whatever this guy knows about fiber, he really is out of his depth when talking about routing.

    I can’t speak for KotL’s points, but I recognise a pattern here. Here’s someone who because of their long and valuable experience is incredibly confident in stating blanket and absolute statements about their own area of expertise, without really conveying that there might be a variety of opinion in their own topic area, and who is also unaware when they are wandering off their own reservation into places where their original positions no longer hold water, or they’re no longer the experts they think they are.

    I’m not saying that those who have such tendencies have opinions that should be rejected absolutely. Personally, I tend to grant them more respect than most, because I’m always impressed by in-depth knowledge (and I’m a sucker for engineer authority figures).

    But I’ve also been burned many many times by supposed experts confidently opining like this, who then turn out to be totally wrong. Examples include managers of IT departments explaining to me how software development works (before being eaten alive by their competition who apparently didn’t), major figures in the telecom industry confidently informing me that no-one would ever need more than 14.4Kbps upload bandwidth, and long-time newspapermen explaining to me in words of one syllable why the Internet was not a threat but Rupert Murdoch’s Delphi was.

    The biggest hint, I’ve found, for this situation, is when these experts call those proposing other ideas “idiots” rather than actually engaging with their points in a reasonable fashion. It’s a giveaway that they’re relying on their own initial reaction of bluster and outrage to drive their arguments.

    As I get older, I watch out for those reactions in myself, too. Usually if my first reaction is “OMG what idiots! They know nothing about teh internet”, I stop and rethink, in case I am about to become a prisoner of my own prejudices. It makes for longer and more boring comments on forums like this, but it helps me find out the truth, which I enjoy.

  19. Danny O’Brien – there is actually a name for this though I am spacing on it right now. Basically it comes from how engineering is taught as an authoritarian discipline where knowledge is assumed to be absolute and complete.

    Takuan – Fon? Seems similar, except that it can be hacked. I’m been thinking of getting one… sometime.

  20. Danny O’Brien:
    What a thoughtful comment! I too know a lot of really really smart people who act like you describe. It is always amusing to see them in a discussion with someone who knows that they are blowing smoke.

    I remember a friend trying to tell another friend what her job entailed. The second friend couldn’t believe the arrogance of the first friend. She proceeded to correct him about what her job actually entailed. Friend one did back down, but had to save face by saying that OTHER people who had her exact same job did it differently. Since I didn’t have “other people” with me to tell him he was still wrong, I let it go.

    I too have an over appreciation for “experts”. To see some truly ridiculous experts see the new movie Changling where experts try and convince a woman that another boy is her son, even though the new boy is 3 inches shorter!

    Who are you going to believe, the expert or your lying eyes?

  21. Very reasoned and intelligent statement, Danny. We’ll be sure to round you up early with the other dangerous ones.

    As to wifi meshes, don’t begin to know enough to comment, but I’m watching all this with avid interest. I take a great comfort in knowing that those far above me in technical understanding in this area are just as much unsure as to what will finally unfold.

  22. Hello everyone,

    Thanks for taking an interest in this paper, and I appreciate the feedback.

    I’d like to address Keeper of the Lantern’s statements. He is right that I am not an engineer, which is why I interviewed several analysts and some actual network operators in the course of writing this paper (they are thanked at the end). Based on my research, a point-to-point or “home run” network — where a dedicated strand of fiber runs from the home to the PoP — is both feasible and advantageous. The alternative, noted by KotL is a “point to multipoint” network, where a single fiber run from the PoP, and, at some point in the field, there is a signal splitter, and from there individual homes are served with individual fibers. Most commonly, this is accomplished via a PON.

    If you want a good overview of different topologies and their costs and benefits, I highly recommend the San Francisco Fiber Feasibility Study cited in the paper:

    KoTL stated: “The article focuses on running fiber all the way from the CO to the home using the “home with a tail” model. That’s … not only infeasible but (most likely) pointlessly and impractically costly.”

    Now, a home run network is clearly feasible. Amsterdam’s CityNet, for instance, is a home run network, as are others in Europe.

    The key benefit of a home run network is that it enables competition between service providers, allowing them to use the layer 2 technologies of their own choosing. As Herman Wagter, CityNet’s Managing Director explains:
    “If you have a strand of fiber to each apartment, … [i]t allows unbundling and competition on a line-by-line basis without duplicating the expensive physical layer. It means that on each customer-end line you could have a different operator. We are saying that we believe totally in competition. This means that in the switch house [central office] or interconnection point (local point of presence) we provide for different racks for different operators, because on a line by line basis customers could sign up for different combinations of offerings.”

    In contrast, a point-to-multipoint PON network does not allow for this sort of competition. There can be only one operator who “lights” the fiber. This is explored further in:
    Anupam Banerjee and Marvin Sirbu, ”Towards Technologically and Competitively Neutral Fiber to the Home (FTTH) Infrastructure,” 2005, (available at:

    Importantly, a home run network is truly future proof. Again, in a home run network like CityNet, individual service providers have the flexibility to provision the network with whatever technology they like – Active Ethernet, PON (with the PON at the PoP), or some other technology. In turn, as network technology advances changes, there will be the flexibility to take advantage of those changes. In contrast, if the PON is buried in the field, the network is locked into that technology, unless they want to rip out the PON, which is costly.

    Further advantages and disadvantages are explored in greater detail on pages 104-105 of the SF study.

    The chief criticism of home run networks is not that they are pointless, but rather that they are too costly. This point is widely debated, and in the paper we concede that it is somewhat contested.

    The key cost is supposedly that it is too difficult to terminate thousands of fibers at the PoP. Interestingly, CityNet terminates over 10,000 fibers at a single PoP without any problem. Indeed, this topology is no different than the traditional copper telephone network, which terminates thousands of fibers at a single PoP, and terminating fiber can be easier.

    In the paper, we cited to Yankee Group analyst Benoit Felten statement that “The difference in cost is limited; there is not a huge gap in costs.”

    Felten explores this further here:

    (I also spoke with both Wagter and Felten separately who provided further details on this front.)

    As we concede in the paper, there are many obstacles to this model’s success. But, there is no strong reason to think that the model is technically impossible, and, in particular, no reason why a home run network is impossible.

    Note: one can also imagine a customer owned network that was in fact point-to-multipoint. The problem is that then all the customers would have to jointly settle on having one service provider to “light” the fiber. As a condition of becoming the active network operator, the customers could then make that operator to provide open access at layer 3 to other competiting service providers. However, this would introduce another collective action problem into the model.

  23. Thanks for that very good response Derek.

    One question: if we want to have a choice of ISP (over our fiber connections) would this be physical or logical? The latter seems kinda cluttery.

  24. JS7A:
    Yes, Palo Alto and other places have had Fiber to the Home for years, but that’s not what’s being proposed here. FTTH is last-mile. Slater and Wu are proposing that it is both feasible and desirable to run fiber (one supposes single mode) all the way from the Central Office to an end user.

    Let me address Derek Slater’s comments by saying what I LIKE about the proposal, and (what I assume) was Cory Doctorow’s initial impetus for posting about.

    I like the idea of opening up the ownership model and allowing individuals and/or communities the option to own (at least) their own last mile or perhaps (and I think this is far less desirable) a set of optical connections all the way to the CO.

    What’s interesting is that this model (if there could be a way to make the economics work) might even be better for some triple-play service providers, though of course it probably means that they can be switched to a competitor (in some cases) far more easily.

    One could also argue that home run fiber is indeed future proof. Given what DWDM already offers commercially, I don’t think anyone in the world (even George Gilder!) would say that conventional single mode fiber won’t be able to support future bandwith needs (at least not within a generation or two). A technological upgrade then becomes a matter of switching out equipment in the CO and the home (rather than a truck roll to replace equipment or infrastructure in the field).

    I get all that: On the surface, kinda utopian and with some advantages.


    (I’ll post my concerns in the next post.)

  25. My real concern with the Slater-Wu article is embodied by their statements around Moore’s law applied to telecom bandwidth and, in particular, to transceiver speeds at the end users.

    Actually, from 1995 to at least 2003 (and most likely until the present day), telecom bandwidth has far outstripped Moore’s law. DWDM (Dense Wavelength Division Multiplexing) has caused the price of telecom bandwidth to come crashing down dramatically (and this was indeed the root cause of many of the big service providers going bust during the telecom bubble). Single backbone fibers that previously supporting 2.5 gig now routinely support 100s of gigabits and even terabits.

    Meanwhile, transceiver speeds have not kept up, but the important point here is that they have not needed to get faster in order to generate all the information that has been pumped into the worlds big fiber optic networks.

    The main reason why faster transceivers do not need to scale with Moore’s law even though the backbones (and bandwidth prices) do, is primarily due to the death of the circuit-switched model.

    Modern packetized traffic a la TCP/IP is bursty in nature. Packetized flows rarely require a constant and fixed bandwidth. Even the heaviest flows will have plenty of moments of near-silence.

    As a result, packetized traffic benefits highly from sharing the medium. Different users of that medium (especially for big pipes) will only very rarely all need all of the bandwidth at the same time. In the rare cases where that happens, the router fabrics will drop packets and then upper layer protocols can retransmit if they choose to. This is the very nature of TCP/IP and the packetized world, and it works pretty well (even supporting VoIP calls in real-time).

    But it is also what allows for broadband buildouts into neighborhoods to be cost effective. This will likely be even more true as we go to higher bit rate buildouts, because why would I want to pay for a 10Gig transceiver (and dedicated fiber) when I will actually only need that full bandwidth very rarely? (In fact, even with the heavy apps of the future you will need only a tiny fraction of this amount.) It makes far better sense for me to statistically share that fiber’s bandwidth with my neighbors and (more importantly), we only need a single 10gig transceiver pair on that fiber. I can access that fiber with some far smaller (and cheaper) transceiver and upgrade (perhaps) if my needs grow.

    Let us point out too that having a dedicated 10gig connection to the home pretty only pushes the problem back to the CO. It’s not like the internet can (or indeed ever will) switch your 10gig connection all the way to the internet server of your choice (even if they were willing to receive it). Your packets are going to get multiplexed ANYWAY, so what has your 10 gig line really done for you?

    As for costs, current broadband buildouts are designed to have payback periods of 2 to even 5 years, and that’s with pretty much all of the various bandwidth sharing tricks thrown in. By building dedicated fiber all the way from the CO to the home, at best we will be pushing to the far end of that payback time, if not moving beyond it.

    Will homeowners be willing to invest in something that won’t pay them back for ten years? If they were, we’d likely see far more deployments of special insulating windows, solar panels, geothermal heating and the like. But consumers are far more frugal than this and will not fall in love with an ownership model that forces them to pay large up-front costs.

    Are there buildout environments where this kind of model can work? Perhaps. But I think that for this idea to become viable, the model will have to consider the upper layers (most particularly layer 2.5, or MPLS), and offer not a fiber-based handoff of an idnvidual’s traffic, but an already packetized and multiplexed handoff to the service provider (ie, 10 gig of neighborhood traffic).

  26. I don’t know anything about, uh, Internet…hardware. There appear to be some sort of physical fibers involved. Something called packets. And lots of mention of having something that goes somewhere called “CO”. In any normal circumstances I would find this post both utterly incomprehensible and horrifically boring. Like, I really can’t overemphasize the extent to which this stuff is out of my subject field.

    Yet for some reason I just read all the comments. With, like, avid interest.


    Maybe it’s because when geek types argue, it’s unusually like sparring. Not real sparring, but movie sparring. With scoffing, laughs of contempt, triumphant cries of “Oho!” and reluctant, dignified concessions. It’s like…Spanish conquistadors. I picture you both wearing fedoras and twisting your moustaches. I know that’s probably really rude and inappropriate to say and I respect you all immensely for your knowledge but there’s something about discourse like this that strikes me as inherently adorable.

    Infinite are the arguments of mages, man!

  27. Actually, from my vantage point letting each home owner pick his own ISP is not a very good idea at all. Letting the homeowners’ association collectively bargain for group discounts is a much better idea, maintaining competition and driving down prices like no individual could. My last homeowners’ association would constantly get cheaper, faster broadband by renegotiating the contract ever two years like this.

    It’s like cable – if the homeowners’ association buys cable in bulk the cost per user is lowered dramatically.

  28. This is a great idea, and has been done several times. For example, Springville Utah has a large community run ISP that brings fiber to their homes called back fence (actually, it’s not so big anymore).

    Where I live (Orem, UT), the fiber network is owned and operated by the city and different carriers lease the lines. It works out pretty well as I get 50Mbit up and down for $50/mo.

Comments are closed.