top bar
QuickTopic free message boards logo
Skip to Messages

TOPIC:

Shirky: Why 3G is doomed

8
Hosidax
02-01-2010
04:20 AM ET (US)
7 years later I can't resist asking...
How's that prediction working out for you Shirky?
7
FlukePerson was signed in when posted
03-31-2003
04:16 AM ET (US)
In the UK we've already seen the demise of one 'nearlynet' - the Rabbit phone network, which at the time when analogue cellphones were first taking off offered a cheaper alternative of a series of access points at popular locations where people would congregate.

Unsurprisingly, people wanted to also receive calls as well as make them and they didn't want to be limited as to where they could use their phones. True, the situation may be different for data but it's an interesting historical precedent.
6
Erik V. OlsonPerson was signed in when posted
03-29-2003
03:49 PM ET (US)
Taking it point by point:

No offense, y'all, but you're all missing the point: for data (not voice), Wi-Fi is better because people typically don't need or want to be streaming multimedia in the cab ride from the airport. They don't need to be downloading PowerPoint presentations or making interactive collaborative documents while striding purposefully down the street.

No. But of their sitting in an office or hotel room or airport, and the only WiFI point that's showing is WEPd, they're screwed. And most people don't even need that much bandwidth. The most important service on the internet, still?

Email.

The whole posit on WiFi eliminating 3G is that WiFi is basically everywhere that you need it. This is *incredibly* not true, despite the number of Starbucks in the world.

3G's voice component is a very miniscule part of the push to 3G. Voice will take such a tiny tiny part of the 3G bandwidth that it'll essentially be free -- if the data infrastructure can pay for it.

Actually, voice is the the driving reason for 3G. 3G lets the entire connection go as packets, and significantly increases a given cell's call handling ability. Well, acutally, it increases the cell's data handling capability -- since, to a 3G cell, it's *all* data.

Wi-Fi doesn't need to be ubiquitous: it just needs to be fast, reliable, and cheap. That's Shirky's point and mine, too. You simply can't afford to build out the same level of performance *per user* in a cell network.

Whenever the cell guys talk about 2 Mbps, they're talking about per available cell channel. They have a very small number of these per tower. They foresee massive increases in data use. That 2 Mbps channel of course would also have voice priority.<I>

First, there's *lots* of 2Mbps channels per cell. Because all the 3G cell does is *pass data*. Most of that data, right now, will be digitized voice on a virtual circuit. But it's all data to the cell. Much like, well, WiFi!

You are thinking that there's no market for a large are 56KB connection -- if you don't have at least a megabit per second, it's unviable. This is, of course, untrue. The *vast* majority of the world connects at 56KB or less.

<I>So when 10 people in the same picocell try to download that PowerPoint presentation and someone streams video, and 500 people are using cell phones...you're back to modem speeds.


You assume modem speeds are unsellable. This is, of course, untrue. You assume that WiFi will always have more bandwidth availble to the user than 3G. This is also untrue, see below.

Because Wi-Fi hot spots are currently designed for data (although voice over IP phones will change that, to be sure) and typically have 256 Kbps to a couple Mbps on their backhaul, you have a different set of expectations. If usage increases at a commercial hot spot, you bring in more bandwidth. This isn't possible with 3G: they can't just buy more spectrum.

First, there's something you need to understand, since you keep missing it.

It's *all data*. It's all bits. Current cell systems are hampered because they tried to keep the voice metaphor going, and use circuts. They switch circuits in clever ways, but they're still radio circuits. 3G joins WiFi in taking circuits out of the low levels. It's all packet data. Doesn't matter if it's a phone call, a movie, or a PGP message. It's bits. In packets. Just like WiFi. Just like the Internet.

Second, if you get ten-twenty users on a hotspot, the bandwidth per drops to modem speeds. Furthermore, if a set of hotspots is connected by one wire (See DFW and MSP airports), then many users are sharing that bandwidth. The problem is greater for cells, but it is, by no means, nonexistent for WiFi.

Wi-Fi isn't attractive because it's free in some places; it's attractive because it's so easy to install, upgrade, and make denser.

It is easy to install, for a single POP, but making denser? Incredibly hard, due to channel limitations, and the fact that even if you very carefully set up your WiFi hubs with sectors and channel offsets to get maximun density, you lose when Joe User plugs in his airport at the office and hammers on a section of your cell -- or, worse, his 2.4Ghz cordless phone. You also have exactly *0* rights to complain about this -- the spectrum is unregulated.

I could wire a coffee shop in Seattle for maybe $1,000 setup and $500 per month to handle 500 simultaneous users. But I could also spend $200 setup and $50 per month to handle 50 simultaneous users.

So, for 500 users, at .5 megabit per, simultaneous, you'll need to provide 250mbits of bandwidth, or they'll start noticing degredation (remember -- you posited that this was unacceptable.) Given Voice-over-IP, streaming video, and the like, the idea of users having continually saturated connections is not at all improbable.

So, first. Throw out that Airport hub. You need much more than the 10mbits it can provide. You either go with 25 802.11b receivers, or 5 802.11a boxes. Except, of course, your range drops with .11a. So, start buying thos 802.11b boxes. Assume cheap peices of crap. That's $2500 right there.

Now, bandwidth. T-1? Not hardly. That ultra-top end DSL line that the vast majority of us can't even get, 6mps? Ain't enough. T-3? Nope -- that's 44mbs. You need 250mbs -- so, a stack of T-3s/OC-3s, or an OC-12. Even if you drop to 56kbits, (basically, a tenth of that .5 mbits, above) -- you still need 25mbits on the backbone supporting those 802.11 points -- That's 1mbit, per backbone. A T-1 or ADSL line would do nicely. You'll need 25. Assuming you don't care about outbound bandwidth (and, if you posit Voice-over-IP, that's a mistake), that's $50 each, per month. $1250. Presuming your DSL provided doesn't choke when he sees the load. Plus, of course, the phone line cost to bring the DSL in. Plus the cost of setup time, and maintenence.

You could hook them all up through ether, then have just one 25mbit wire. But that means all 25 have to be within ether range limits, and you have to buy that 25 mbit wire.

$500 a month? Dream on. Never mind the reliability issues with cheap hardware, the routing issues with that dense a network and other fun stuff.

It's scalable, it's cheap, and it's available in the right places -- or soon will be.

Except, of course, it isn't, scalable is *never* cheap, and unless the right place is in San Fransciso or Manhattan, it's not available.

Tell me how many SFO passengers will use 3G services, even at a couple megabits per second, on a *metered rate* when they can pay, say, $8 for up to 24 hours of unlimited multiple T1 usage over Wi-Fi?

Well, few. Unless you're in ORD, where, unless you're in the right Airline Club, you can't pay any dollars for WiFi. Or, if your hotel doesn't have WiFi. Or wherever.

The world's *much* bigger than SFO -- or even San Francisco. WiFi's great, if you've got a connection that you can use. Easy in some airports -- harder in the real world, and in many cities, there's no choice. Pick up your phone.

In Airports, where Joe User can't easily start broadcasting on 2.4Ghz, it can be done. In dense urban enviroments, it becomes much less practical. The WiFi revolution has only worked so far because there aren't that many WiFi hubs contesting for space yet. Note that a private WiFi hub, using Wep, is one that helps NearlyNet none -- but eats a channel from NearlyNet.
Edited 03-29-2003 03:50 PM
5
Glenn FleishmanPerson was signed in when posted
03-29-2003
02:24 PM ET (US)
No offense, y'all, but you're all missing the point: for data (not voice), Wi-Fi is better because people typically don't need or want to be streaming multimedia in the cab ride from the airport. They don't need to be downloading PowerPoint presentations or making interactive collaborative documents while striding purposefully down the street.

3G's voice component is a very miniscule part of the push to 3G. Voice will take such a tiny tiny part of the 3G bandwidth that it'll essentially be free -- if the data infrastructure can pay for it.

Wi-Fi doesn't need to be ubiquitous: it just needs to be fast, reliable, and cheap. That's Shirky's point and mine, too. You simply can't afford to build out the same level of performance *per user* in a cell network. Whenever the cell guys talk about 2 Mbps, they're talking about per available cell channel. They have a very small number of these per tower. They foresee massive increases in data use. That 2 Mbps channel of course would also have voice priority.

So when 10 people in the same picocell try to download that PowerPoint presentation and someone streams video, and 500 people are using cell phones...you're back to modem speeds.

Because Wi-Fi hot spots are currently designed for data (although voice over IP phones will change that, to be sure) and typically have 256 Kbps to a couple Mbps on their backhaul, you have a different set of expectations. If usage increases at a commercial hot spot, you bring in more bandwidth. This isn't possible with 3G: they can't just buy more spectrum.

Wi-Fi isn't attractive because it's free in some places; it's attractive because it's so easy to install, upgrade, and make denser. I could wire a coffee shop in Seattle for maybe $1,000 setup and $500 per month to handle 500 simultaneous users. But I could also spend $200 setup and $50 per month to handle 50 simultaneous users.

It's scalable, it's cheap, and it's available in the right places -- or soon will be.

Tell me how many SFO passengers will use 3G services, even at a couple megabits per second, on a *metered rate* when they can pay, say, $8 for up to 24 hours of unlimited multiple T1 usage over Wi-Fi?
4
Mark AllertonPerson was signed in when posted
03-29-2003
01:05 PM ET (US)
Even if we accept the logic of permanets and nearlynets, I believe Shirky is misclassifying both 3G and WiFi on the basis that the classification criteria for data and voice are different. If we apply the same criteria as for voice, then clearly 3G is a "nearlynet" in the same way that cellphone service is, and WiFi is a "barelynet", like landlines - and is technically incapable of achieving "nearlynet" status. We already know how nearlynet vs. barelynet plays out. Is it really safe to assume that people will not find compelling uses for data in the same places that cellphone usage is compelling?
3
smithdm3Person was signed in when posted
03-29-2003
02:05 AM ET (US)
Another reason why 3G likely won't fail is that a large part of the infrastructure deployed is also required to boost the capacity of the cell network for voice as well. Hence it won't suffer the same problems as the airphone because it has a number one use. As the voice becomes packetized and moves away from a circuit switched network it becomes easier to carry data.
2
Erik V. OlsonPerson was signed in when posted
03-29-2003
12:07 AM ET (US)
More: Why airphones are failing. Yes, they aren't getting the use, partially because of the enormous rates. But this is mere marketing stupidity. Businessmen, esp. in a recession, won't pay $2.50 a minute for calls. Would they pay $.50 a minute? In a heartbeat -- hell, I'd pay a couple bucks to be able to drop an alert from a plane.

Right now, WiFi is attactive -- because a few kind people, and a great deal of incompetent people, are broadcasting nearlynet for free. Counting on this to last is a sucker's bet.

I like WiFi. I use it. I ocassionaly set it up when I travel. But I'd pay a reasonable amount of cash for a 36kbit connection that I knew would *always be there.* WiFi isn't. Outside of a few cities, WiFi is the exception, not the rule.

There are a great deal of people who might need to be on the net, right now. WiFi and NearlyNet can't give us that -- and won't give us that, by design.
1
Erik V. OlsonPerson was signed in when posted
03-28-2003
11:55 PM ET (US)
I don't buy it -- yet. WiFi's certainly cheaper to deploy, but pi*r^2 means that it's cheaper to deploy 3G cell towers than the dozens of WiFi hubs needed to cover a given area -- and cells, at 900mhz, have a much easier time with obstructions than WiFi at 2.4ghz (or, worse, 5GHZ)

The reason for the WiFi explosion is cost -- hubs are cheap, wire to the internet-as-large is cheap. If WiFi could get better spectrum and/or better power (say, 1W at 400mhz), then, boy howdy, would WiFi sweep the globe. (Note, however, there's lots of stuff in the low UHF band. However, I think we could shave a few UHF TV channels away, now that the part of the tech world that needs dozens of channels uses cable or satellite to get them.)

WiFi's getting the press -- but cell providers go for levels of reliability that hackers and .coms only dream of. Does your commerical WiFi setup have redudant transmitters, antennas, power sources and connections to the net?

Print | RSS Views: 1017 (Unique: 608 ) / Subscribers: 1 | What's this?