802.11 Task Group Updateby Glenn Fleishman
An alphabet soup of standards is about to overflow the boundaries of committee meetings and fill in the cracks in wireless networking. The IEEE 802.11 Working Group has been busy over the past year, moving several lettered task groups closer to ratification of standards for speed, scheduling, security, frequency coexistence, and other important matters.
In an article I wrote last summer, New Wireless Standards Challenge 802.11b, I covered the various proposals and specifications in process, and how they might affect wireless networking. In the meantime, although specifications haven't been finalized, the industry has moved awfully close to key compromises and consensus on the next-generation of several important elements that should offer greater speed, reliability, streaming voice and audio/video, and a baseline security that one could actually rely on.
Task Group G: High Bit Rate
Our dear friend 802.11b is about to accelerate into the future in the backwards-compatible 802.11g specification. Task Group G was dedicated to finding a higher bit rate (22 Mbps or more) in the 2.4 GHz band where b already operates, but without abandoning b and its installed base.
The 802.11a spec, which operates in the 5 GHz band at a raw speed of 54 Mbps, was seen as a competitor and colleague at the same time to b and g. It competes in that it's faster, but it is collegiate in that innovations in its area can be applied to 2.4 GHz as well.
Chipmaker Intersil proposed the OFDM (Orthogonal Frequency Division Multiplexing) that they developed as an encoding method for 802.11a be brought over to 802.11g, as well.
This nearly caused a Balkanization of the spec, as their competitor, Texas Instruments, believed so strongly in their PBCC (Packet Binary Convolution Coding) technology that they had started to make equipment incorporating a 22-Mbps version of PBCC. Their systems would be backwards compatible with 802.11b, but would not operate with 802.11g equipment when it was released, unless the final g spec had a backwards-compatible mode.
In November, a last-minute compromise was reached. 802.11g would support all 802.11b encodings for complete backwards compatibility. It would also use the plain version of OFDM included with 802.11a. These two encodings would be mandatory.
As an option, manufacturers can use a modified version of OFDM or Texas Instruments' PBCC. This means you won't find g radios that can't talk to each other, but you may not be able to eke all the performance out of devices that have chosen different optional encodings. This is more like a divided Berlin than Balkanization.
This compromise was approved, but g as a whole isn't scheduled to be ratified until the second half of 2002. This hasn't stopped companies from developing g-based equipment; chipmakers are releasing samples and reference design in mid-2002, with the expectation that manufacturers would ship hardware by the fall.
Look carefully at "final draft proposal" g devices: make sure there's a guarantee for upgrade or replacement when the spec is ratified, otherwise you might have to replace equipment.
Task Group A: 5 GHz, 54 Mbps
The 802.11a specification is in full swing with several manufacturers shipping equipment based on Atheros chipsets, and other companies coming online shortly with their own chip offerings.
A debate has sprung up in the IT community, played out in the pages of InfoWorld and elsewhere, about whether 802.11a makes sense today, as opposed to waiting for future integration of other 802.11 specifications.
It's clear that the 5 GHz band is less crowded; there's practically nothing in it. The 802.11a spec offers more channels and a wider band, so you can have more access points with less interference and clearer signals; these signals can offer overlapping coverage in the same physical area.
But it's equally clear that without cheap client radios in cards and various PCI form factors, 802.11a won't have anything but niche adoption, especially in scenarios where server-level network performance is needed and physical wiring isn't cost-effective or available.
At a wireless conference last November, I heard a lot of talk from chipmakers and others that dual-band radios will emerge first in client devices, like PC Cards, allowing a single client to hop onto the 802.11g/b bandwagon and the 802.11a service (although almost certainly only one at a time).
The IEEE carefully ensured that everything except the radio part of 802.11a, b, and g were compatible. A chipmaker could supply two radios without duplicating the rest of the necessary circuitry, keeping costs fairly low.
Also, because of European regulations, it's probably most worthwhile to wait for 802.11h's innovations being folded into 802.11a (see below). 802.11h requires new radio equipment, meaning that existing 802.11a devices cannot be upgraded to incorporate that additional standard.
Task Group H: Spectrum Managed 802.11a
Unlike Ethernet, which Wi-Fi is a kissing cousin to, 802.11 radio specs don't listen before transmitting to see that the line is clear. Instead, they transmit and without getting appropriate feedback, halt and retransmit. (Ethernet devices listen, send, and if they sense problems, wait a random amount of time to try to retransmit.)
802.11h overlays 802.11a to solve both interference and overuse problems, as well as improve coexistence with other specs that might reside on the same band. The h spec requires devices to check whether given frequencies are in use before transmitting (Dynamic Frequency Selection or DFS), as well as only transmitting at the minimum necessary power level (Transmit Power Control or TPC).
These additions were formulated specifically to meet requirements for using the 5 GHz band in the European Union, which has been promoting its own specification called HiperLAN2. 802.11a coupled with 802.11h meets these requirements, offering the chance for 802.11a to bull its way into a market that many had thought was already lost to an incompatible format -- which still hadn't appeared in shipping devices.
Devices using 802.11h should be out before the end of 2002.
There's a chance for spillover of h into other standards like b and g, of course, to improve their responsiveness. When you read the notes from the IEEE proceedings, you see that this is clearly in the minds of the members developing the specification.
Task Group E: Quality of Service
Every packet has an equal chance of getting through in 802.11b. Task Group E wants to change that, allowing for what's known as "quality of service" or QoS, to guarantee that some packets have more priority than others. This is a fairly tricky task, involving coordination between client radios, access points, and system administrators.
QoS is needed for consistent voice-quality calls using VOIP (voice over IP) and for streaming multimedia. 802.11e could overlay several different IEEE standards, including b, g, and a.
The advantage that HomeRF, another 2.4 GHz wireless LAN spec, had always offered against 802.11b was its built-in scheduling requirement: voice packets were given priority on a regular interval to ensure that a HomeRF phone handset, or many handsets in simultaneous use, wouldn't cut in and out on the talker.
802.11e should be approved during 2002, possibly in the next few months. (Several IEEE 802.11 votes were postponed because of Sept. 11, delaying specs that were expected to be approved or further along in 2001.)
Task Group I: Enhanced Security
Originally, 802.11e covered both scheduling and security. With the constant release of weakness reports in the WEP (Wireless Equivalent Privacy) encryption system built into 802.11b, however, security pupped into its own group, letter I.
Task Group I has been working to find a replacement for WEP that, hopefully, would also have enough compatibility to be implemented without vastly revising the current generation of systems.
The group's meeting minutes indicate a high level of concern about not abandoning existing users. The task group approved a measure late last year to ensure that a backwards-compatible, interim security improvement would be issued before the group's work was done.
The proposal that emerged a few weeks ago was a method to improve current WEP technology by simply creating a larger number of initialization vectors for WEP encryption. Even though WEP keys are labeled 64 and 128 bits, 24 of those bits were essentially set to one of a few known values. To make it worse, other aspects of WEP make it relatively simple to capture a few hundred megabytes of data and build a table from which the WEP key could be extracted.
The interim proposal just ups the exponential scale without radically revising the system, turning WEP back into a reasonable first line of defense that would require possibly 100,000 or more years to crack with current technology. (Other weaknesses might again surface reducing that to 15 minutes.)
The long-term goal of 802.11i, however, is to replace WEP. The failure in public confidence has the group looking at specifications that are at a much higher level of complexity but still computationally efficient enough to embed in lower-power, inexpensive devices, such as chipsets used for PC cards.
The failure of WEP resulted in the group dropping the name WEP2 for the new standard and replacing it with Temporal Key Integrity Protocol (TKIP), something which is much more descriptive: assuring that a key retains its security over a period of time. The ultimate standard will almost certainly involve rotating through many keys over short periods of time.
To complicate matters, the more general 802.1x protocol (not 802.11x as is sometimes reported) has to be taken into account by Task Group I. 802.1x is a method of authenticating users through a back-end system in a secure fashion. Some weaknesses in the approach have already been discovered, unfortunately, as there is a lot of room for man-in-the-middle style interception.
The bottom line: with no set date for approval of a new keying specification, and the weaknesses in 802.1x, it may be 2003 before an entirely new keying system is approved by the IEEE and incorporated in devices. In the meantime, the interim proposal may find itself in firmware upgrades throughout 2002.
Look for the Wi-Fi Label?
While standards development has progressed at the IEEE, the Wireless Ethernet Compatibility Alliance (WECA) -- the group that owns the Wi-Fi trademark -- has continued to push forward on incorporating various new specs into the Wi-Fi brand.
The head of WECA told me last November that WECA has plans to only update its certification process for checking that devices are Wi-Fi compliant once a year. For instance, in late 2001, they added "ad hoc" mode compliance to their specification; this is the mode in which computers can talk peer-to-peer without routing or an access point. Before their certification update, each major chipmaker had their own standard for ad hoc mode.
The Wi-Fi brand will extend into these new realms as additional IEEE specs are ratified and released. But it's likely we'll see as much as a one-year lag between some of the current IEEE task groups finalizing their work and inclusion in the Wi-Fi process. And because they want to add a single change per year, it's possible that security and quality of service won't be added simultaneously without special action by the WECA board and members.
WECA also added a new mark for testing 802.11a: Wi-Fi5 (for 5 GHz). It's already being bandied about early in 2002, even though WECA doesn't plan to formally launch it and their certification process until spring 2002.
Putting it All Together
Yes, it's possible that by the end of 2002 you could buy the alphabet soup of all devices: a PC Card that said, "802.11a, b, e, g, h, and i compliant."
And that would be a good thing.
Glenn Fleishman is a freelance technology journalist contributing regularly to The New York Times, The Seattle Times, Macworld magazine, and InfoWorld. He maintains a wireless weblog at wifinetnews.com.
Return to the Wireless DevCenter.