G-whizz clouds the issues

We all want a faster home network but preferably don't want to rewire the house. Yet strangely almost nobody, except Broadcom, seems pleased with the way the politics of wireless networks are going with the new 802.11g standard.

This was intended to speed up standard 2.4GHz 802.11b transmission to the point where it would carry five times more data - a rated 54Mbits/sec. But unfortunately everybody wanted to be first to market.

The result is that, although the new chips work, they may be disappointing. In particular, it looks as if anybody who sets up an 11g network where there are 11b users, will see a 30 per cent speed drop.

It might be a lot worse and it might not be that bad, depending on how active the 11b devices are.

One Linksys executive, who works with all the major chip manufacturers and designers, said: "Believe me, they were all desperate."

His company sells most of the home-networking wireless products in the world and, in his view, at least three major chip designers wanted to jump the gun and be the first with the new 54Mbit products - even if this broke the standard.

Broadcom hit the tape first and, at the time of going to press, looks like it has managed to ship close on 200,000 devices of what it calls '54G' to distinguish it from the IEEE 802.11g standard.

At the time Broadcom froze its design, it was confident that the standard would not change between the preliminary specification and the final approval.

There was one problem, which it thought was insoluble, caused by the fact that the 11g modulation scheme - orthogonal frequency division multiplexing - is incompatible with the spread-spectrum technology used in 11b.

So whenever an 11b packet is transmitted, the whole network has to slow down to 11b speeds.

However, Texas Instruments had cracked the problem using technology it developed for a non-standard 22Mbps product of its own and has now adapted for 802.11g.

I had considered that when both 11g and 11b cards share an access point all devices would have to operate at 11b speeds. However, Texas Instruments engineer Bill Carney told me I was wrong.

"It is understandable how this can be concluded. It is based on your experience with products that use the faulty, pre-11g implementation of Broadcom's 54g chipset," he explained.

Carney pointed out that Broadcom made a key mistake in its interpretation of the draft standard on how a mixed mode network was to behave.

He wrote: "In fact, the IEEE 802.11g [committee] spent numerous hours discussing the ways to preserve the performance advantages of the higher-rate devices in presence of legacy 11b."

A number of mechanisms were included in the draft specification to allow this, according to Carney.

"It's unfortunate that Broadcom's incorrect implementation is now setting a market expectation that you must upgrade your entire network to 11g, or suffer the performance fallback to existing speeds," he said.

Other chip vendors, having interpreted the draft specification correctly, offer performance advantages for 11g nodes in a mixed network, explained Carney.

"Because of this and other interoperation issues, the 54g technology is quickly being dropped in favour of other solutions. In this case, being first to market with a pre-standard offering ended up doing more harm than good," he said.

This is powerful stuff - so uncompromising that I checked with Texas Instruments' PR department. They made it clear that they endorsed Carney's viewpoint, and added the information that he was "very senior" in the Wi-Fi Alliance hierarchy.

When I posted this story it was as if I had prodded a hornet's nest with a red-hot iron. The problem may be related to the fact that in (secret) compatibility tests, there was some indication that not all rival 11g chip sets will actually work with each other.

Whatever the reason, it seems that the senior members of the Wi-Fi Alliance attacked Carney for speaking out of turn at a time when the controversy was supposed to be resolved behind closed doors.

Carney now says that he was quoted "out of context" but hasn't yet explained what, if anything, was incorrect, apart from the fact that he was not speaking as a member of the Alliance.

Frankly, he should be; what's more, he should be more explicit in his attack. To be running compatibility tests after the release of commercial products is an extraordinary situation for the industry to find itself in.

Worse, few assurances are offered about 'what happens if I buy an 11g product now, and it turns out to be obsolete when the standard is ratified?'

The answer from Linksys, for example, is: "We'll ship upgrades free", which is fair enough. But when you ask what happens if it turns out to be a hardware problem, the only answer I've been given is: "It won't be."

A source (not Carney) inside Texas Instruments indicated that this may not be the case but that, anyway, the need to come up with a solution that doesn't make 54G obsolete will certainly delay the ratification.

"It may not be available now until after July, and maybe after October," said this source.

It's understandable that the Wifi Alliance should be touchy about this. I've seen threats that "nobody in the Alliance will ever talk to you again, Guy" simply because the Alliance has been caught asleep at its post.

The purpose of the Alliance is simply to ensure that products aren't sold as Wi-Fi unless they are tested for compatibility by the Alliance.

Broadcom's decision to jump the gun drove a steamroller over the guard house, crushing everybody involved and leaving them feeling very small and vulnerable.

They have to do something clever - preventing public discussion may well save their public image. But will it solve the problem?

Like this story? Spread the news by clicking below:

 del.icio.us     Digg this     reddit!