Receive new posts as email.
This site operates as an independent editorial operation. Advertising, sponsorships, and other non-editorial materials represent the opinions and messages of their respective origins, and not of the site operator. Part of the FM Tech advertising network.
Entire site and all contents except otherwise noted © Copyright 2001-2010 by Glenn Fleishman. Some images ©2006 Jupiterimages Corporation. All rights reserved. Please contact us for reprint rights. Linking is, of course, free and encouraged.
Nothing interesting happened in 2009: The top 10 most visited pages at Wi-Fi Networking News in 2009 were stories from 2003 to 2008. Go figure! In fact, only 13 of the top 50 viewed pages were stories from this year, although about 15 percent of page views were visits to the site's home page.
Topping the list is that old favorite, a white paper on security: Weakness in Passphrase Choice in WPA Interface.
The site had about half a million itinerant visitors who came and left, and about 80,000 routine visitors who read on some regular basis via the Web, email, and RSS.
I thank all of you for continuing to read the site and provide comments, feedback, and news. I wish everyone the happiest of new years.
I must note that this article about chicken wire was written by Geoffrey Fowler. Seriously.
(Chicken wire background by Elné Burgers; used by Creative Commons license.)
Dana Spiegel posts an exchange with a writer at the Heartland Institute, wherein the writer starts with a bias and then stoops to insults: Oh, lordy, our friends at the industry-funded (but-we-won't-disclose-who) Heartland Institute are trying to explain how evil municipally funded free Wi-Fi is again. I thought I was back in 2005, again.
Given that Miami Beach and St. Cloud, Flor., are now just about the only free city-wide Wi-Fi funded by a city in the US, I don't see the urgency in Heartland trying to explain why it's an evil entitlement.
The Heartland "reporter" (hrmph) tells Dana Spiegel that in St. Cloud "the City Council tried to shut down their free WiFi service because of the expense but stopped that initiative after residents who could not pay for their own Internet access protested loudly. In that vein, do you foresee municipal WiFi networks like Miami Beach’s at all becoming another entitlement program for Americans?"
That's some myopia you've got there! St. Cloud residents of all stripes, not just those who "could not pay" for Internet service--that's bought-and-paid-for thinktank code for "poor and probably African American because they're poor"--were interested in keeping the service alive. The city council and mayor responded to the outpouring of interest, and funded the network further as a result.
We last tangled with the Heartland Institute in depth in 2005. I wrote a piece called "Sock Puppets of Industry" (1 February 2005) that spelled out undisclosed funding and other conflicts in a report issued on municipal broadband that was riddled with errors.
The Heartland Institute continues to claim that it is "not affiliated with any political part, business, or foundation," which continues to beggar my imagination as a statement. The institute has concocted ever more elaborate explanations as to why it doesn't disclose donors, which are known to include major firms in industries about which the institute produces anti-regulation reports. Corporations and foundations provide 89 percent of funding (per the institute's 2009 report), with corporations making up 13 to 16 percent. The foundations are the usual suspects that push money to thinktanks to create research and reports that allows affiliated companies and institutions to cite data as independent of the funders.
All you need to know about Heartland may be encompassed in the Sock Puppets post, in which I note that (at the time) Heartland had a Philip Morris executive on its board, even as it wrote reports and a book denying peer-reviewed scientific and economic analysis of the societal costs of smoking. In the comments, Heartland's head denied that such an executive was on the board, despite the fact that the Web site listed that executive and his affiliation.
Temporary TSA rules in the wake of attempted detonation of something during Detroit landing mean no electronics during last hour of incoming international flight to the US: The rules in place as of today for flights arriving in the U.S. prohibit anything in your lap, standing up, using the restroom, and any access to your carry-on baggage. Carry-on limits have already been tightened as well, further than the existing tiny ones. The last-hour limits sound similar but more restrictive than the rules when flying into Reagan National in D.C.
These rules don't yet appear to apply during the first part of a flight, but it's typically at least 20 minutes after takeoff until the 10,000-foot limit has been passed, and later when the fasten seatbelt light is turned off.
So far, the rules apply only to arriving international flights, but the New York Times notes, "It was not clear how often the rule would affect domestic flights." Earlier today, it sounded as if some domestic flights would also see this restriction.
For in-flight Internet providers Aircell and Row 44, this could be devastating if the rules cover domestic flights or flights that spend significant time over the US.
For flights up to 3 hours, which I believe are at least half of U.S. domestic mainline (non-regional jet) flights because of the hub system, that leaves precious little time to use any electronics, and makes it unlikely someone will pay the $10 rate for those flights. For longer flights, the $13 rate Aircell levies may be more reasonable, but travelers will now be mentally removing 90 or more minutes from the useful time they have.
If these rules remain in place, I will be curious how quickly Aircell and its partner airlines change pricing, and whether in-fight Internet upgrades to planes continue at the pace they were at.
Connexion by Boeing had a lot of factors against it, but one of the key ones was new rules and a drop in airline revenue following 9/11.
[This post was updated later on Saturday to make clear that it appears only international flights entering the US would have these final-hour restrictions, but it remains up in the air, literally, as to whether those rules will also affect domestic flights.]
Update: My pal Xeni Jardin at BoingBoing had the same idea after flying back from Guatemala yesterday.
Ford offers a remarkable option to add Internet access to an auto: Unlike automakers who have signed up to build in the Autonet Mobile system, a Wi-Fi gateway with a built-in cell modem, Ford is opting for a bit of openness. Car owners with the next generation of Ford Sync (by Microsoft) coming next year can plug in their own USB broadband modems into a slot near the gear shift to enable Internet access over Wi-Fi to passengers. USB modems are offered by all carriers, including Virgin Mobile's currently unique pay-as-you-go plan.
Ford should also be praised for not just offering, but requiring WPA2 Personal security. You might want to share access with others, but not after you see your first cell data bill with overage charges on it.
By allowing a USB adapter, Ford lets a car owner who already has a 3G subscription plug in instead of adding yet another subscription fee on top.
Sync costs $395 as an optional add on, although it's included in certain high-end Ford and Mercury models. The system is designed to allow integrated use of cellular phones and digital media players, as well as provide emergency service and send car system status reports. Next year, turn-by-turn navigation will be added (three years at no cost).
Connectify releases Windows 7 PAN enabler: Connectify has pushed out the 1.0 release (free!) of its software that turns on a hidden feature in Windows 7--never completed with the proper user interface--to allow a single Wi-Fi connection to accept local connections while itself connected to a Wi-Fi network (infrastructure style). This avoids the trouble of ad hoc networking, while allowing robust WPA2 security.
Cleveland considers 4 1/2 sq mi free network: What's most interesting about this plan may be the proposed cost: $600,000 to build. In the olden days, it was a few hundred thousand dollars per square mile, as I recollect, even though it was often billed as "$100K," but that didn't include the modern density that's understood to be needed, and the real bill for the back end. Cleveland has other areas with free service through the One Community effort that sprang out of initiatives at Case Western University.
Bluetooth's low-energy mode announced: As part of Bluetooth 4.0 (even though 3.0 is just starting to ship now), the low-energy mode will provide networking for sensors that can't carry huge or rechargeable battery packs. This will be useful in healthcare, alarm monitoring, fitness, and other categories. The ZigBee standard was supposed to eat up this kind of usage, being a low-power, low-bandwidth technology, but Bluetooth wants to sweep this use inside its existing ecosystem. The data rate will be 1 Mbps.
Running through last week's top stories: Verizon extends free Wi-Fi at thousands of US hotspots to 3G laptop/adapter users (Windows only, thanks), Continental signs up for 21 planes with Aircell, and McDonald's ups the ante in its competition with Starbucks by unleashing free Wi-Fi starting mid-January in the United States.
Jaunted tells you which, what, how much: The always fun travel site Jaunted has done a superb job documenting which U.S. airlines have service, what you pay on each, and a good summary of which planes in the fleet are enabled.
The site separately summarized this information even further into a chart. If you're trying to get Internet access while flying, bookmark that page.
It goes around and around and comes out there: The city of Philadelphia has announced its intention to purchase the Wi-Fi network from Network Acquisition Company (NAC), a firm that itself acquired EarthLink's in-progress network for a song with a promise to build it out and to change its name from the placeholder it chose. Apparently, the placeholder turned out to be correct: the firm acquired the network and operated it, but it seems little else emerged in its plans, made before the massive economic downturn. NAC took control of the network in June 2008 (see "Eleventh Hour Rescue for Phila. Network.")
The Philadelphia Business Journal seems to have but sketchy details about the deal, which would commit the city to spending $17m from 2011 to 2015 (fiscal years) to expanding its core fiber network and integrating and expanding the Wi-Fi network. The wireless network would be used for municipal and public safety purposes, as well as limited public place Internet. Phila. told me years ago that it spent millions each year on leased digital lines from telecom; many cities have built fiber networks and rings to conserve that cash in house while boosting network speeds often by a factor of 10 to 100 times the leased line rate.
Update: The Philadelphia Inquirer has more information. The city will pay NAC $2m, which is roughly the same amount that NAC paid EarthLink and other parties. The $2m from the city will comprise $1.5m from homeland security grants and $500K from public-safety funds.
In this case, the city claims a $9m cost conservation against $17m in spending; the operating savings don't include increased productivity or other measurable improvements outside of pure network operation costs, however.
This is a far cry from Philadelphia's 2004 plan to give free Internet service to everyone via Wi-Fi; EarthLink's goal for Wi-Fi at subsidized and dial-up prices to residences through outdoor transmitters; and NAC's plan to mix free, fee, and business services of varying kinds to make a go of it.
Philadelphia is now trodding the path that many other cities have followed in the last five years, which is focusing on government efficiency through cost conservation and using Wi-Fi and public safety wireless as an adjunct to core wired networks.
Continental Airlines opts for in-flight Internet, and chooses Aircell's Gogo service: The initial commitment is to put the service on 21 Boeing 757-300s, a fleet that serves cross-country routes, and seems like a good test deployment. Aircell never discusses its financial arrangements with carriers; it's possible that the firm is subsidizing a pilot installation in order to prime getting Continental's mainline fleet equipped. There's just no way to know.
With a deal signed with US Airways, but no plan released for deployment, Aircell has locked up all the major US airlines plus Air Canada, with competitor Row 44 having signed just Southwest and Alaska. Those deployments are apparently on hold until Row 44 raises sufficient capital to proceed.
Starting in January 2010, US McDonald's restaurants will no longer charge for Wi-Fi: The quick-service chain has service in 11,000 of its 14,000 locations in the US, and previously charge $2.95 for two hours access. The network is operated by AT&T, originally contracted by Wayport which was acquired by the phone giant last year. Most AT&T subscribers get free access at McDonald's, Starbucks, and a few thousand other locations.
This might drive more traffic to McDonald's, which is fighting for various new kinds of customers with its specialty coffee service. Starbucks offers two hours daily of no-cost service as a reward to regular customers--a reward program that changes later this month to require several purchases before then getting the free Wi-Fi deal. (See "Starbucks Makes It Harder to Get Free Wi-Fi," 3 November 2009.)
A McDonald's executive quoted in the Wall Street Journal is clearly making a Starbucks taunt when he says, "free is free." The no-cost service starts in mid-January 2010.
Will this change the perception of the value of Wi-Fi in the US? It's hard to tell. AT&T provides its service to over 27m subscribers now, which is a decent subset of all users, and Starbucks must have millions of regular users. I pay $10/mo to Boingo Wireless for unlimited access at lots of different venues, and $10 has been more than worthwhile to avoid paying $3 or $4 or $6 a pop for access as I need it.
Google certainly gave for-fee Wi-Fi a kick in the keister by sponsoring so much free airport Wi-Fi this holiday season. That may make it hard for airports to go back to charging, but there may be more sponsorship or other models in the future. An increasing number of larger airports have moved to free service (ad supported, often), or are considering such a move. With particular ad models that require a substantial user commitment, such as watching a 30-second commercial, that may pay off airports: get 5 to 10 times the number of users, and ad revenue could cover costs, as well as be a well-received amenity that makes travel better.
Certainly, McDonald's offering free Wi-Fi puts pressure on anyone who doesn't, whether you like the chain's food or not.
Verizon's limited, irritating Wi-Fi access for broadband subscribers now extends to mobile broadband customers, too: Old! Unimproved! But extended! Verizon's Windows-only, laptop/netbook-only free Wi-Fi access, enabled via a subset of Boingo Wireless's network, has been extended to its mobile broadband subscribers. That's laptop adapter/card 3G users--not smartphone users, who still don't get free Wi-Fi.
Not only do you have to use Windows 2000, XP, Vista, or 7, but you have to run Verizon's VZAccess Manager, and manually select a network when one is available. No manual login, no Mac OS X support, no shoes, no service.
I wrote about the limits of Verizon's deal back on 28 July 2009 in "Verizon Limits Free Wi-Fi to Laptops." Verizon is only offering access to "thousands" of hotspots nationwide.
This is in contrast to AT&T's deal, which now encompasses 27m subscribers to DSL, fiber, or business services, as well as all the laptop 3G users, and all iPhone, BlackBerry, and Windows Mobile smartphones with Wi-Fi built in. Smartphone subscribers can only use the free service from the smartphone, which is an automatic connection on the iPhone and some other models. Non-smartphone subscribers can log in with a user name and password from any device; no special software required. And AT&T gives access to its 20,000-strong hotspot network. (That's less of an issue come January when the 11,000 McDonald's stores in that network go free.)
My friend John Moe interviewed me for Future Tense on MPR about airport Wi-Fi: John and I talked about how airports suddenly seem to have free Internet access all over, why that happened, what's happening with the biggest airports, and shutting the lid on your laptop sometimes while you're traveling!
Trying something new: Would a weekly round up in a minute or two of wireless news be of interest? Listen to this week's podcast: 2 1/2 minutes, three stories. Comments welcome. Use the player below, or you can download an MP3 or AAC file.
The WiGig Alliance has hit a planned mark, with a spec going to members and 7 Gbps in the air: The WiGig Alliance is a group now comprising about 30 members, including all major wireless chipmakers, that wants a standard radio platform and standard application profiles for the 60 GHz millimeter-wave band. The 60 GHz band is available with different allotments in most international regulatory domains, with some, like the US, having 7 GHz available. It works best within a single room due to high attenuation from physical objects.
The WiGig Alliance is attempting to avoid the travails of standardized UWB, which took several years to fail at the IEEE, and then more to arrive in scattered fragments in a market that doesn't care, as well as the delays that encumbered 802.11n.
The announcement today is that the group expanded its membership over the last several months, and finalized the first version of its spec, which will be handed over to members to review, and then released likely in the first quarter of 2010. The spec will hit 7 Gbps of raw throughput per channel, up from about 6 Gbps in an earlier draft. As many as 4 or 5 channels will be available in any given space.
The goal is to produce a single radio standard with flexibility, allowing both high-performance and low-power devices that can work interoperably on the same band; and a set of application profiles or purposes, to allow video, data, and other kinds of transmissions to work without being at cross purposes, require separate chips, or emerge from disparate trade groups.
This is contrast to both the IEEE 802.11 Task Group ad (802.11ad), which is developing a WLAN protocol for 60 GHz, and Wireless HD, a trade group led by SiBeam with some overlap with WiGig's membership, and focused entirely on high-definition in-room streaming.
"60 GHz is a whole clean sheet of paper to work on, and to use it only to replace a single wire seems to be a tremendous waste," said Mark Grodzinsky, a board member and the marketing head at Wilocity. There's a lot of things that can happen over it." The group hasn't optimized "so heavily for one particular usage at the cost of others," he said.
Key to the spec's development was the desire to have devices that operate at lower power and consequently low data rates work interoperably on networks and with other devices that are using the maximum data rate, potentially for high-def streaming.
Grodzinsky said that beamforming, for instance, in which multiple antennas are used to steer a signal, could include many antennas for a high-throughput device, fewer for a low-power device, and none at all for an "ultra low power device."
While video may be on everyone's mind, the group has developed a spec that is entirely backwards compatible with existing Wi-Fi standards at the MAC level, including security. With chipmakers deeply involved in WiGig, this could mean WLAN adapters would have 2.4, 5, and 60 GHz radios, and move interchangeably among them based on power, range, and other characteristics.
Ali Sadri, the chair and president of the WiGig Alliance, and the WPAN/60 GHz standards director at Intel's Mobile Wireless Division, said it was critical to have a single specification in place around which all manufacturers could rally.
"90 percent of the Wi-Fi chipsets are being built by the members of the WiGig silicon team," he noted, which could make it easy to gain traction as an extension to Wi-Fi.
Grodzinsky said that the 802.11n standards battle taught everyone many lessons. "We'd like to think that we can learn from our mistakes," he said, noting that there's "no point in being fierce competitors" for technology that doesn't exist.
Many IEEE members belong to firms involved in WiGiG, and it's likely that 802.11ad will be shaped by proposals coming from those groups.
The biggest risk may be devices that share a radio standard but have disparate capabilities, something that the Wi-Fi Alliance faces every day as more protocols and features are added. "There will be a way from a consumer standpoint to recognize exactly what you're buying," said Grodzinsky.
This piece in the Globe and Mail may be a little too breathless by half, but it's quite fun: The first two planes are plying routes from Canada to the US with Gogo Inflight Internet enabled over American portions, although this is a trial and only until 29 January. The service is being offered commercially. The Canadian part will be enabled in a partnership between Canuck firm SkySurf, which won the domestic airwave rights, and Aircell, which operates Gogo.
The WHDI Consortium has finished a spec for running 1080p at 60 GHz and 12 bits over 5 GHz spectrum: It's not Wi-Fi, nor anything remotely like it, but the WHDI spec uses 40 MHz channels in the 5 GHz band to carry the equivalent of 3 Gbps as far as 100 feet. This new spec, based on work from Amimon, which developed the technology, boosts resolution fro 720p in the previous version to 1080p. It also supports HDCP, the digital rights management (DRM) specification that's used with wired HDMI to ensure end-to-end protection of content.
I wrote about the several contending wireless high-definition specifications in contention back in February 2009 for Ars Technica, including this detailed explanation of how the WHDI/Amimon system works.
While 5 GHz Wi-Fi using 40 MHz channels can only claim with 2x2 MIMO and two data streams to deliver a raw data rate of 300 Mbps, WHDI will deliver 3 Gbps. Because they aren't.
From the press release: the new spec "supports the delivery of equivalent video data rates of
up to 3Gbps." That's equivalent.
WHDI uses a clever system of representing visually more important data in the encoding such that it's more likely to get through in the worst circumstances. Ever less significant information is encoded in methods that are ever more susceptible to interface. The more noise, the less insignificant information gets through.
But that's where you have that equivalent: during the best transmission times, the WHDI Consortium's spec will be able to push through what looks like uncompressed 1080p; during the worst, something far lower than that. It's unclear whether best efforts will win the day.
The competitor for WHDI is pretty clearly WirelessHD, backed by SiBeam, which uses 60 GHz millimeter-wave signals for as much as several Gbps (real Gbps, not equivalent) for each of several channels. The 60 GHz signals are limited to within one room, and have some non-line-of-site and obstruction issues; the 5 GHz service can work over longer distances, but the WHDI probably doesn't want to go into any great depth on how rapidly a signal degrades providing an equivalent bandwidth that's far below 3 Gbps.
The WHDI Consortium was set up by Amimon, Hitachi, Motorola, Samsung, Sharp, Sony, and LG Electronics. Those are some pretty big names in the consumer electronics space. The WirelessHD group also, oddly, includes Samsung, Sony, and LG, as well as Philips, Intel, NEC, and Toshiba, and a host of chipmakers including Intel and Broadcom.
A clever researcher launches a service that mostly proves how easy it is to crack a bad WPA password: IDG News Service reports that Moxie Marlinspike, a nom de Net for a security researcher, has launched WPA Cracker, a $34 service that cracks poorly chosen passwords found in a database of 135 million passwords in 20 minutes or less using distributed commodity computing. For $17, it takes 40 minutes.
Let me be clear: this is a clever and worthwhile addition to penetration testing (pentesting) and network security, and I would gladly pay $34 to prove to someone smug that his or her company password was vulnerable. But it is not a generic nor dangerous attack on WPA. Smart companies, likely millions of them, already use account-based network authentication in the form of WPA/WPA2 Enterprise, which is not vulnerable to this form of brute-force attack. WPA/WPA2 server-side support is de rigeur in the enterprise network infrastructure, and available from third parties, as well as built into Microsoft Server and Mac OS X Server operating systems. Home users and small-business users are most likely to employ simple passwords.
The ability to crack WPA-PSK (Pre-Shared Key) passwords used in WPA/WPA2 Personal has been well understood since the IEEE 802.11i committee was first nailing down the details of how a password plus other material would be transmuted into a master key used for encryption. Research Robert Moskowitz let me publish his paper, "Weakness in Passphrase Choice in WPA Interface," way back on 4 November 2003, because he wanted to highlight how Wi-Fi equipment makers were letting down users by not guiding them to pick a strong password. Routers now typically come with configuration software that encourages picking a good password and changing the default network name. (That article is still one of the most-read pages on my site six years later.)
To be crackable, keys have to be both short, typically eight or fewer characters, and comprise words found in dictionaries, including alternatives for those words with common substitutions, like a 0 (zero) for the letter O or 3 for E. Longer passwords and ones that use combinations of letters, numbers, and punctuation will remain uncrackable over potentially very long periods of time. WPA cracking becomes something close to exponentially more difficult with each additional letter added to a password. (A generic WPA-PSK crack could change that near-exponential factor, but it hasn't appeared yet.)
As Marlinspike notes in the site's FAQ, a WPA-PSK uses the network's SSID or name as one component, and the network password as the other. The SSID is a salt, which is ostensibly random information added to a password to make extracting the password more difficult. That works quite well in cases in which the salt can't be easily determined; in fact, randomizing the salt for per-packet encryption was one of TKIP's big improvements over WEP.
However, for Wi-Fi, because the SSID is sent in the clear or can be sniffed during an association, someone needs just capture a four-way handshake that's part of the WPA/WPA2 Personal protocol; that handshake can be provoked by sending a disassociation request, too. The disassociation request and handshake take just a moment. A cracker could also monitor a network for such requests, too, which can happen quite frequently as machines rejoin the network.
The 20-minute time is the duration for salting each of 135 million passwords--which Marlinspike says in the FAQ are tuned to be likely ones used for this sort of purpose--and checking the results against the captured handshake. WPA Cracker charges the same $17 or $34 for a recovered password or the answer that the password isn't in the set.
While Marlinspike wouldn't tell IDG which cloud-computing servers he was using, it might be easy to figure out if you do the math. 400 Amazon cloud-computing units of its "standard on-demand instances" running Linux costs precisely $34 per computing hour. That's a little too close to be an accident. Amazon provides an API to allow launching instances with custom made OS images, and charges a full hour at a time, rounding up any partial hours. Thus, Marlinspike breaks even on a single queued job, but sequentially queue work in which the average cracking time is under 20 minutes will produce profit. The researcher can have 400 units fire up and then shut down all while he is sleeping.
You can download precomputed sets of common passwords salted with common SSIDs, but Marlinspike notes in the FAQ that these publicly available sets are limited to 1,000 SSIDs and a million words per set. Someone who enables encryption is likely to change the SSID as well.
Over at Ars Technica, I write about how Wi-Fi is getting goosed for the future: It's not all about speed. That's the key message I kept hearing from people who develop and work with 802.11 networks. Rather, future flavors of Wi-Fi will combine aspects of higher throughput, better system capacity (more devices across a node), robustness (removing dead spots), and resiliency (better dealing with network congestion and interference).
It's true that the IEEE is working push 802.11 past 1 Gbps, but that will be likely via 80 MHz or even 160 MHz channels, possible only in 5 GHz and in certain circumstances. Having a greater number of streams per access point may have more impact, by not just improving potential speed, but extending range and filling in coverage holes.
Apple's October revision to its AirPort Extreme Base Station and Time Capsule lets these units run at up to 450 Mbps, 50 percent faster: I knew that Apple had put a 3x3 antenna array into these devices, which could be argued was intended to improve speed-over-range, a common reason to add antennas. (Read the background in my 20 October 2009 article, "Apple Slipstreams 3x3 into Wi-Fi Base Stations.")
But I had heard that Apple had built three streams in, making these devices capable of a raw 450 Mbps operation, or 50 percent faster than nearly everything else on the market. Each stream in 802.11n can carry more data than 802.11g--more than twice as much, in fact, when wide (40 MHz) channels are used. Apple makes wide channels available only in 5 GHz, which isn't unusual; 2.4 GHz is quite crowded and full of competing uses. Thus 450 Mbps is the raw rate in 5 GHz, 225 Mbps in 2.4 GHz.
How do I know this when Apple won't reply to calls and emails on the topic? I was tipped to it by several wireless industry folks who wanted to remain anonymous--nobody likes to offend Apple--but didn't have a definitive label until today.
I regularly check the Wi-Fi Alliance's site for new certifications from Apple, because more information often appears there than the details Apple provides. Sure enough, based on a query from a colleague tonight, I did a search and found that on 3 December 2009, Apple's two Wi-Fi gateways had been labeled with three-stream support for transmit and receive in 2.4 GHz and 5 GHz.
[Update: The inestimable Tim Higgins of SmallNetBuilder wrote in to note that he had discovered, disassembled, and documented all this a month ago! I'm sorry to have missed it. Read his article in which he also found that Apple had connected a couple of antennas incorrectly. He confirms that it's a Marvell radio; Marvell started shipping three-stream chips nearly two years ago.]
This aggregate of 675 Mbps of raw data (something closer to 400 Mbps of TCP/IP and over 300 Mbps of net throughput) between the two simultaneously available bands is still theoretical: Apple's and most manufacturers' Wi-Fi adapters still are only sport 2x2 antenna arrays with two-stream configurations. (Apple updated its AirPort Extreme certification on 30 November to reach full 802.11n compatibility, too, but only for two streams.)
It's already been asked, but gear from before October with 2x2 arrays cannot be updated to 3x3; it's a hardware change, as is the jump from two streams to three streams.
In the short run, Apple base station buyers will get the benefit of greater range and greater speed over short ranges than with older gear. In the long run, though, there's a big speed boost to come.