Andy Seybold and Ron Sege (Tropos) hammer away on metropolitan-scale Wi-Fi: I've had long internal debates with myself about how to write about this issue played out in competing guest commentaries on Muniwireless.com. Andy Seybold is a respected figure in the industry, and someone I admire. But his approach to external Wi-Fi, however reasonable some of his concerns are, has been ham-handed, often inaccurate, and biased towards licensed frequencies.
Because he's a consultant and does not have a list of his and his firms' clients, it's impossible to know what angle he comes at this. I'm not suggesting his opinion is paid for. He's too honest, too independent, and too smart for that. But if you just had your head inside the cell data helmet for two years, metro-scale Wi-Fi looks absurd. Take off that helmet, and evaluate it fairly, and you could have an entirely different take. I'd urge Seybold to disclose any past and present consulting arrangement with companies that compete in the space that he is offering public opinion about. He's not a journalist, but he still writes like one.
His opponent in this debate, Ron Sege, makes his money as the CEO of Tropos Networks, a company that is the leader in selling metro-scale Wi-Fi mesh equipment. So we know where his bias is: he'd like his company to sell more and more gear. He has every interest in making his approach seem workable. But he's also responsible to his private shareholders and board of directors as well as his customers. As recent years have shown, pretending something works doesn't work as a long-term business strategy.
(Me, I accept advertising through third parties and am not involved in negotiating or signing advertisers to my sites. I work as a journalist, primarily, and do not consult in this or any industry.)
The difference between Seybold and Sege is that Sege can give you the names and addresses of networks and city IT managers: you can go and try his networks and talk to the people running it who aren't responsible to Sege, but to taxpayers and city officials. Seybold is poking holes through what I have to say is often specious or inaccurate reasoning; Sege is offering a rational approach that's not overhyping the abilities of the system he sells. I think both parties would agree that the future for metro-scale wireless (not Wi-Fi) is extremely bright.
If you view metro-scale Wi-Fi as a poor cousin to cell data, then I have to say that's where the drugs have kicked in and you're channeling Hunter S. Thompson. Verizon Wireless keeps making bizarre statements about how their EVDO service works everywhere unlike Wi-Fi which works mainly when your laptop is physically touching an access point. Okay, I'm exaggerating. But their statements have been strangely broad especially when their technology provider, Qualcomm, has a campus-wide Wi-Fi network that they're very happy with. Seybold agrees: indoor deployments of Wi-Fi are great uses of the technology and they work.
EVDO is fantastic technology that I'm in love with, but let's remember three salient points: limited spectrum available for 3G in this country; high cost for unlimited usage to deter too many subscribers; limited bandwidth compared to the backhaul capable with modern Wi-Fi (mesh or fixed hotspot or hotzone).
So where's the dispute? Let me start drilling into Seybold's Muniwireless.com commentary. He hates 2.4 GHz: it's a messy band. It may experience a tragedy of the commons. It's like Citizens Band radio: too many users turned CB into something no one can use. (Except that it's still in use by a group that carved their own purpose out of it when the FCC walked away.)
But that's not what's happening in 2.4 GHz. The band has become more and more useful because it employs technology to allow many simultaneous networks to work without rendering each other useless. Yes, the more networks, the worse performance. But I've been at trade shows--Wi-Fi Planet, notably--with hundreds of 2.4 GHz Wi-Fi networks over a few thousand square feet, and you can still associate and send data. The FCC hasn't walked away: they're actively involved in tweaking and enforcing rules. Seybold claims companies are selling gear that flaunts Part 15. Hey, who are they? Let's report them. They're violating the law and threatening public safety and corporate data networks through their gear.
Seybold moves on to airports, indoor spaces that you think he would admire. But a lack of coordinated policy have doomed some of his connections, he says. I and others asked where in the comments, and he cited Dallas/Ft. Worth and San Jose as having several networks that apparently prevented him from getting a good connection. But those two airports have coverage from Wayport, which he doesn't mention as one of the signals he saw. I was recently in Seattle and Austin's airports, which are two of Wayport's oldest installations, and had great service throughout. As you imagine, I have professional interest in wandering around to look at signal strength and throughput. I saw other networks, sure, but the ones that Seybold cites are ones that are designed to cover small areas, like an airport lounge. If you're not in the lounge, you might see the signal, but the coverage shouldn't be good. This is frustrating for T-Mobile HotSpot subscribers who aren't lounge members, but that doesn't mean that Wi-Fi failed them.
Seybold's airport reasoning is conclusion by anecdote. Airports are generating hundreds of thousands of Wi-Fi connections each month. Ask Concourse, T-Mobile, and Wayport, to name the biggest players. If service were as poor as Seybold maintains, this wouldn't be happening. I had terrific results in Seattle, Denver, and Austin a few weeks ago, three of the oldest Wi-Fi'd airports in the country. (Seybold is also incorrect about a remark in the comments to his commentary: "access points are being deployed without knowledge or consent of the airport commission--and sometimes with their consent." The FCC ruling last June precludes airport authorities from restricting unlicensed wireless.)
The commentary devolves into speculation about how metro-scale Wi-Fi networks can't work well because of interference and many competing networks (home and otherwise), and how if they even manage to work now they will fail in the future because of a tragedy of the commons.
Unfortunately, all developments point otherwise. Seybold mentions the 5 GHz band in passing, but it's clear that as 2.4 GHz becomes more crowded--I completely agree it will--that the 23 channels in 5 GHz for relatively unused 54 Mbps communications today and 100 to 600 Mbps communications with 802.11n in 2006-2007 will take up the slack. Manufacturers are clearly moving towards integrated dual-band chips in all non-consumer devices. It doesn't cost much more at this point, and it's the way the enterprise is moving.
Combine that technology direction with the spatial multiplexing and multipath discrimination that will appear in 802.11n (and is already in early form in MIMO gear hitting the market), and you solve another problem. If you can more clearly differentiate signals as they reflect in complex, radio-crowded environments, then you effectively increase the amount of bandwidth available across a given geographic area in a given slice of spectrum.
Thus even if 2.4 GHz becomes unusable due to crowding with today's technology, tomorrow's technology won't be subject to the same limitations. Even better, you can continue having bad results with today's technology while tomorrow's is installed all around you. Tropos could move from 802.11a/g to 802.11n for backhaul and use multiple radios for service to support legacy users.
Seybold also writes, "The problem with 2.4-GHz Wi-Fi is that if it works in a given wide area today, there is no guarantee that it will continue to work tomorrow. Building a system that requires, for example, 500 access points today might require the addition of another few hundred access points in the future. This would throw a wrench into the business model."
That's a lot of different ideas, but I don't buy any of them. The technology will improve, so upgrades to the technology will be necessary. But all of the plans I've seen and read about involve the idea that technology will improve. A 500-node network that needs 200 to 300 more because of usage or other factors is already in the plan. Nobody is deploying a network of fixed size, crossing their fingers, and trusting that it will work indefinitely--or even 1 to 2 years in the future without adding nodes.
Seybold transition into questions of mobility, or accessing metro Wi-Fi while in motion. "If public safety officers have to pull over to the curb to run a license plate while they are in pursuit of a vehicle, what good is the network?" I don't think Seybold has talked to police officers about how they work to make that statement. Most of the selling point of public-safety networks is about keeping staff in the field instead of returning to base to fill out paperwork. Another part is about getting robust information in the field--but not, typically, at 100 mph pursuit. You're probably on the radio at that point and focusing on driving and not getting shot rather than typing on a keyboard (or having your partner do such).
In any case, focusing on mobility sells the idea that a technology that doesn't yet exist in most cities--broadband speed cell data, which is coming--and that requires payments to external providers trumps a flexible, multi-purpose network that a city itself could own or have built for it. Cities should probably think about conserving costs in areas in which outside providers have no similar interest. This is one of the primary problems in my view with state laws that would prevent municipalities from being able to build multi-purpose networks that public safety personnel would benefit from.
Like so many of the arguments in this commentary and more cellular-focused articles and chats elsewhere, Seybold wants to make the indirect case that an unlicensed band will devolve into chaos without rules that provide for strict separation of providers, cell-like seamless handoff, and other features common to cellular data networks.
But he's taking a very small slice and a set of strawman that I don't think hold up to scrutiny to posit that today's networks don't work (when they do) and that the same technology will get worse and worse instead of the inevitable path that's already underway to improved use of spectrum, better signal discrimination, and more channels for use overall.
Now you think I have forgotten about Tropos CEO Ron Sege's commentary on Seybold's piece? I have not. Here's my dilemma. I'm not a toady, but I agree with practically everything Sege writes. Why? Because he's not trying to create an reductio ad absurdum argument. Sege is willing to consider and even introduce points of view contrary to his own interest in the purpose of arriving at a logical conclusion.
Sege doesn't look as Seybold does at spectrum in the classical, early 20th century view that is being widely discredited by people as varied as open-source radio enthusiast and the FCC. Spectrum is only scarce when you spew radio waves over it. It's abundant when devices are smart enough to use the least signal, to avoid stepping on others, and to hop away from frequencies in use. Some of this is already in place in 2.4 GHz; some in European rules for 5 GHz.
In the non-scarce spectrum worldview, the more transmitters, the more difficult but not unsolvable the problem becomes. Coordination happens among devices using protocols that allow this to be sorted out.
If you apply Sege's arguments to the tragedy of the commons you get a very different outcome from Seybold's. Seybold would argue that in a space intended for 1,000 cows consuming regularly that he found 5,000 cows and the field was trampled. Sege, in contrast, would point out that there were 5,000 cows, but they were led in and out on a rata system that assured that no more than 1,000 cows--and often only a few hundred cows--were munching at every given time.
In fact, rather than 1,000 cows mostly owned by Verimoo or SBCow, the 5,000 cows were owned by hundreds of different dairy farmers. By keeping the commons open and using a protocol that determined the number of cows that could contend for grass, the commons continued to flourish. To follow Sege's commentary, he would say that Seybold didn't stoop to look at the grass at all, but reasoned that 5,000 cows were an untenable number for the commons, and vowed to return in a year to see if any grass was left at all.
Sege's summary is rather stirring and in accord with my opinion: "Cautionary projections of potential failures of technology solutions based on previous failures have a place in the debate, as long as they are fully verified as still valid and acknowledge real changes in the environment."
Comments welcome below that advance a civil discussion of these issues.
Being in the public safety industry for the past 18 years, wireless connectivity in field operations is a vital part of policing our communities. The debate over spectrum and bandwidth has raged ever since the first "radio cars" were deployed and Dick Tracy was fantasy. As far as public safety goes, police have always been able to perform duties in spite of the lack of technology. Law enforcement is typically 5-10 years behind the public sector when it comes to wireless or technology tools for field force operations and thus must continually struggle to implement new technologies. Recent budget crunches have caused technology deployments to focus on keeping the officers in the field and maximizing personnel time and cost.
Now, on the debate over 2.4Ghz. failing or not. The same was said about 400Mhz. UHF, VHF and later 800Mhz. except, all of the private spectrum systems built to push data over 800Mhz. had a VERY difficult time unless text only information was optimized. Many states have spent countless millions (Ohio has an 800Mhz. network for voice and data called MARCS that is in excess of $300M)on private infrastructure that simply becomes unusable after 7-10 years. I believe that the part of the debate which is not being addressed is one of management and economy. When we deploy a police department with wireless in the field, I push many of them to public infrastructure over cellular bands. Why? Although there is a recurring cost, the network and equipment will continue to be upgraded regularly. The responsibility for maintenance on the network falls with the carrier / provider. In cases where "government" takes control over operations, often times the SLA is much harder to manage and outages, failed equipment and continual costs cause the network to break down operationally. In part, these problems arise from sheer buracracy but also suffer from non-intelligent gear. As the gear becomes spectrum aware, network aware, and somewhat self managing, these issues begin to dissolve. Everyone remembers the early days of networking. It took a real engineer to make things work and specialized knowledge to problem solve effectively. Now, networks are self aware and equipment can seek out "best connections". What a difference in 10 years. So, to your point that gear will play a major role, I completely agree. I do however feel that unless a governmental body can satisfy a user base as would a private company (driven by profit and performance), we will continue to see failed efforts but not for the reasons cited about 2.4Ghz. being over subscribed. Failure will be by our own inability to manage resources effectively.
Glenn, maybe you can answer me this. Ron talks a fair bit about how CSMA (collision sense multiple access) in WiFi prevents the tragedy of the commons from occurring, but I'm not sure he is fairly positioning the problem. Isn't he just talking about how CSMA manages traffic between users on a single network, when the problem is more likely to be multiple networks:
If Wayport puts up a single AP at a hotspot, and 100 people are actively using it, and there are no other APs in the area, I understand that CSMA will avoid packet collisions, and allocate a quasi-fair and efficient use of the shared resource.
However, what about when T-Mobile powers up an AP in the same area, on the same channel? The way I understand it, interference between the T-Mobile radios and the Wayport radios are NOT managed with CSMA, meaning that a T-Mobile user's radio will check with the T-Mob AP, get the CTS, and spout packets with no concern of whether it interferes with the Wayport AP. This is why conferences with a managed, single WiFi network can offer acceptable services even with many users, but conferences and shows with many WiFi networks have pitiful WiFi performance. Seriously, can you answer me this?
Also, Ron correctly states that there are power limits on WiFi gear, but let's not forget that Tropos has already taken RF power escalation to the legal limit. A Tropos node transmits at 4 Watts, or over 20X what the average home AP output. What happens to my WiFi network when Tropos puts an AP on my lamppost and on my channel? Can I still see my network from my front room? Sure, I know how to change my AP channel, but not everyone does, and should we have to?
Lastly, lots of Muni WiFi subscribers cannot use standard WiFi cards or Centrino, because the average CPE has about 50mW of transmit power, which isn't enough to be useful on the muni mesh. Chaska, Minn, for example, provides many subscribers with 200mW CPE so they can punch up the signal. I'm not saying this doesn't work, but let's not argue that Muni WiFi has no CPE costs because everyone already has WiFi, when in fact customers need special CPE. Some Chaska WiFi subs also need to put an external antenna either outside their house or mounted to a window.
So there it is, one question and a couple of points. I think both sides of the debate tend to gloss over some of the realities. Oh, yeah, disclosure. I've consulted for both cellcos and WiFi chipmakers, and am shilling for neither.
(I hope this reply looks better after I post it than it looks in the preview!)
Derek asks, "Glenn, maybe you can answer me this. Ron talks a fair bit about how CSMA (collision sense multiple access) in WiFi prevents the tragedy of the commons from occurring, but I'm not sure he is fairly positioning the problem. Isn't he just talking about how CSMA manages traffic between users on a single network, when the problem is more likely to be multiple networks..."
CSMA operates without regard to the "network name" (SSID) or number (BSSID).
Derek also asks, "However, what about when T-Mobile powers up an AP in the same area, on the same channel? The way I understand it, interference between the T-Mobile radios and the Wayport radios are NOT managed with CSMA, meaning that a T-Mobile user's radio will check with the T-Mob AP, get the CTS, and spout packets with no concern of whether it interferes with the Wayport AP. This is why conferences with a managed, single WiFi network can offer acceptable services even with many users, but conferences and shows with many WiFi networks have pitiful WiFi performance. Seriously, can you answer me this?"
By "many WiFi networks," do you mean many providers, or many APs? In any events, the two cases are indistinguishable to CSMA 802.11.
RTS/CTS is strictly optional. Using it adds enormous overhead. It may cause more harm than good. Do those networks at conferences really use it?
Derek, "Also, Ron correctly states that there are power limits on WiFi gear, but let's not forget that Tropos has already taken RF power escalation to the legal limit. A Tropos node transmits at 4 Watts, or over 20X what the average home AP output. What happens to my WiFi network when Tropos puts an AP on my lamppost and on my channel? Can I still see my network from my front room? Sure, I know how to change my AP channel, but not everyone does, and should we have to?"
Ron implies that Tropos uses transmit power control. Let us hope the Tropos AP's receive sensitivity matches its transmit power---that is, it isn't a "crocodile," all mouth and no ears. If so, then CSMA will help your AP to share the medium with Tropos' AP. (Let us also hope the AP on your lamppost uses just 5mW power if that is all it needs.)
I hate to answer your question about changing the channel with a question, but I'm going to anyway. :-) Why should the user should have to change the channel? Consumers and product reviewers should hold the AP manufacturer responsible for automating this. After all, you and I cannot "hear" RF, but our APs can. It's about time they were programmed to tune to a channel where their operation is not impaired.
Also Derek says, "Lastly, lots of Muni WiFi subscribers cannot use standard WiFi cards or Centrino, because the average CPE has about 50mW of transmit power, which isn't enough to be useful on the muni mesh. Chaska, Minn, for example, provides many subscribers with 200mW CPE so they can punch up the signal. I'm not saying this doesn't work, but let's not argue that Muni WiFi has no CPE costs because everyone already has WiFi, when in fact customers need special CPE. Some Chaska WiFi subs also need to put an external antenna either outside their house or mounted to a window."
Clearly everybody does not already have WiFi---did somebody really say that?
I think it is too bad if providers are "punching up" their subscribers to 200mW, since most of the 200mW adapters out there use the highest power regardless of the AP's proximity. It's a waste. I think it is reasonable for the FCC to require transmit power control for the high-power cards.