Columnists

CES: IEEE 802.11n: The vendor-neutering of a once-promising standard

802.11n, as currently implemented in draft form, is incredible technology. It could have been even better if vendors would have exhibited more marketing restraint.

By Brian Dipert, Senior Technical Editor -- EDN, 1/8/2007 11:05:00 AM

This is a commentary that I've wanted to craft for a long time. CES is the perfect venue to finally put my aspirations into action. I'll begin with a brief history lesson. Hang with me, there's a point.

Some of you broadband-tethered young whippersnappers may not believe this, but there was a time when nearly all home-based computer users (with the possible exception of folks like Bill Gates) accessed the Internet over conventional telephone lines. Does anyone remember fetching email over a 2400-bps connection, or via an acoustic coupler? I do (I still have occasional nightmares about it.…Well, not really). As this Wikipedia description neatly details, the history of analog-modem development has consisted of four major phases (with minor perturbations along the way):

1. Vendor-to-vendor incompatibility, driven by vendor-proprietary technology (predated in this particular case by single-vendor de facto compatibility, due to AT&T's line monopoly, which ended with the 1967 Supreme Court Carterfone decision),
2. Multivendor compatibility through a number of generations of increasingly speedy and feature-packed industry-standard formats defined by the CCITT (now known as the ITU-T),
3. A temporary format splintering, as two camps (the Rockwell-led K56flex and US Robotics-championed X2), grappled in the standards meetings and the marketplace for the right to be crowned 56-kbps downstream champion, and
4. A compromise 56-kbps format, V.90, followed by a return to standards body-defined compatibility in V.92.

I still vividly recall swapping out ROMs in my US Robotics X2 modem, in the days before the flash-memory-enabled firmware upgrade era, to provide it with full V.90 compatibility.

Today's mess

The maturation of Wi-Fi (specifically the 2.4-GHz variants) has followed an eerily similar trajectory, and unfortunately I believe we're in the midst of the messy Stage 3 in the process. As my Reality Check article from May of last year documented, wireless network protocols were initially proprietary and numerous; standardization occurred with the Lucent WaveLAN-derived and IEEE-defined 802.11b. Next came the higher-speed 802.11g, whose jaw-dropping and well-documented success requires no further elucidation here.

Up to this point, the various chip vendors and their systems partners worked together, albeit often begrudgingly, in order to cultivate the blossoming Wi-Fi market through interoperability. The first success- and profit-driven cracks in the façade appeared when various 802.11g "Turbo" offerings, such as Atheros' Super G and Broadcom's Xpress, entered the marketplace. "Backward-compatible" with conventional 802.11g, in and of itself a controversial topic, these "Turbo" chip-based devices would deliver higher speed when mated with comparable gadgets on the other end of the Wi-Fi link, the vendors claimed.

The next step in the splintering of Wi-Fi was more substantial, with the unveiling of Belkin's "Pre-N" product line in August 2004. In retrospect, Belkin's naming scheme was more accurate than anything that's followed it; the company made it clear through labeling that its Airgo line of MIMO-based, 802.11g-backward-compatible products was being released prior to the publication of the first 802.11n draft specification one year ago.

Unfortunately, Belkin's marketing move emboldened other OEMs, who have aggressively rolled out routers; access points; streaming-media adapters; Ethernet bridges; PCMCIA-, ExpressCard-, USB-, PCI-, and PCI Express-based client adapters; and other devices based on pre- and post-draft v1.0 802.11n silicon from multiple suppliers. For the last year, I've been collecting and analyzing the field evaluations of these products, and shaking my head with dismay at the consistent conclusions. Not only do products based on different silicon supplier foundations not interoperate, but also platforms based on different silicon-engine revisions from the same silicon supplier, and even those employing the same silicon chip but using different firmware revisions, fail the essential interoperability test.

Heated debates, such as this recent one between Cisco Systems and Meru Networks, showcase one root cause of this incompatibility. There's more than enough wiggle room in finalized specifications, let alone draft ones, to not only allow vendors to craft proprietary "standards-based" products but also for those products to clobber their rivals' performance. Windows Vista's problems with some 802.11g access points' protocols are another example of incompatibilities within supposedly standards-bounded technologies. Didn't someone once say, "The wonderful thing about standards is that there are so many of them to choose from"?

My recently-published cover story began with the statement, "Continuous technological evolution is the inevitable consequence of vendors' desires to sustain revenue and profit at healthy levels." That observation isn't, of course, solely restricted to display interfaces, it's also the fundamental reason why companies are prematurely and, in my opinion, recklessly, going down the Draft 802.11n path. The Wi-Fi Alliance, in the press release announcing its controversial August 2006 decision to certify products based on the upcoming v2 802.11n draft specifications, pointed out that "tens of millions of pre-standard devices will ship in 2007."

A number of leading computer manufacturers either are now or will soon be selling systems containing draft 802.11n silicon, Apple among them. Intel is bundling draft 802.11n support in its next-generation Santa Rosa mobile PC chip set, now sampling and due to ramp into production later this year. Perhaps the to-date pinnacle of 802.11n absurdity occurred a month back when, in conjunction with the announcement of Qualcomm's pending acquisition of Airgo, the company proudly proclaimed its latest chip's compatibility with v2 of the 802.11n draft specification, a document which hasn't even been published yet.

And unfortunately, as I guesstimated would happen way back when this mess began, any pretense of "Pre-N" or "Draft" is largely being dropped from documentation and promotional materials. I've been a CES Innovations Award judge the past three years, so I get an early preview of products that'll be unveiled at the show. Rarely did I see an 802.11n-inclusive product description qualified with a "draft" moniker.

And believe me, this show is flooded with 802.11n draft-inclusive products. The official show directory is 660 pages thick and chock-full of exhibitor ads. The companion directory addendum and guide contain, respectively, 96 and 164 pages and are similarly ad-adorned. None of the advertisements for 802.11n-touting widgets include the "draft" qualifier. I suspect I'll sadly come to the same conclusion as I wander the show floor the next few days. And if "draft" is being discarded here, you gotta believe it's not part of the sales pitch at your neighborhood consumer-electronics retail outlet.

That's the core of my despondency about how 802.11n has evolved. Lots of folks are worried about long-term incompatibility, but I'm not one of them. Granted, the situation will be mighty messy for a long time. To that point and ironically, the vendors now pushing draft 802.11n gear versus comparatively unprofitable 802.11g products are only shooting themselves in the foot. The inevitable technical-support queries and product returns will gobble up any profits they might otherwise garner.

Unlike the analog-modem days, though, the vast majority of equipment nowadays is DSP-based and firmware-upgradeable. Slowly but surely, software up-revs will squash the bugs that currently hinder draft-802.11n compatibility. But the plethora of draft 802.11n silicon already in the market will unfairly constrain the IEEE standards committee as it works toward spec finalization. Unless committee participants are really gutsy (in which case their dictates will probably be ignored by the technology implementers, anyway), market forces will remove from consideration anything that's unsupportable by currently shipping hardware in conjunction with a firmware tweak. Current-hardware compatiblity with the final standard is a particular concern for companies that have already guaranteed that this will happen and would face either backlash (notably in the form of probable lawsuits) or crippling hardware-retrofit expenses if they can't later deliver on the promise.

According to the hard-working IEEE committee, the 802.11n specification is at least a year away from being wrapped up. But as mandated by the marketplace, it's already, essentially, a done deal. Don't get me wrong; 802.11n, as currently implemented in draft form, is incredible technology. It potentially could have been even more incredible, if the silicon and systems vendors would have exhibited more restraint. Alas, that didn't happen, and the end result will be the worse for their cash-driven carelessness. I agree with Maury, let's get 802.11n wrapped up as quickly as it can be. But when it is, I'll mourn for at least a moment what it could have been.