Standard judgement

You're buying a notebook PC, but aren't sure whether to ask for one with Bluetooth, built-in GSM or Wi-Fi. Almost certainly, you're worrying about bandwidth.

Bandwidth is a context. Take backup, for example. Backing up the data on a notebook PC with a 20GB hard disk takes time. If the only connection available is an ISDN line, you probably wouldn't do it.

You'd save some spreadsheets and documents, perhaps, but judging by the way most notebook users complain when their hard disks crash, even those connected to a 100Mbit/s Ethernet cable seem to find the chore of backing up too hard. But 20GB at 128Kbit/s? Never. It's definitely a bandwidth problem. You'd solve the backup issue a different way.

So it's easy to see why network administrators aren't interested in Bluetooth links for PCs. At the maximum Bluetooth is specified for around 1Mbit/s.

Anybody with any experience of wireless connections will tell you the actual data payload is never going to reach that speed - you'll see 600Kbit/s at best in terms of actual data, shifted from one place to another. And if there's radio interference, it'll drop further.

In fact, since the discovery that Bluetooth does, indeed, interfere with wireless LANs using the 802.11b standard, it has to be accepted that it will give you 128Kbit/s, reliably. That's roughly home ISDN speed.

But 128Kbit/s is plenty for someone playing MP3 music. It's fine for sending and receiving emails. It's even acceptable for Web surfing. Actually, these days very few host sites will feed the data to you faster - whatever your connection speed.

Now, suppose that you're sitting on a train with your notebook PC and you need to check your office diary. Next to you is a perfectly ordinary mobile phone with an infrared link to your PC. You dial-in and after spending most of a minute waiting while the two ends chat to each other about training and protocols, you look at the entries for the day. This can all be done in less than another minute - with a 28Kbit/s connection, at best.

Here, the issue isn't bandwidth, but response time. The delay while the diary data comes down the line is invisible; it's the tedious time that it takes the two ends to negotiate a connection that rankles.

By the same token, a PC on a WLAN at 11Mbit/s is still too slow for most people to attempt a full backup of the data. The result is that when people discuss bandwidth issues, they're mostly discussing things that don't matter.

For example, my backup is constantly up to date. I write my text onto a hard disk, but when it's saved, the data is replicated exactly onto a remote Web server. Since I can't write more than a few kilobytes of original text at a time, the data is never held up by bandwidth - even at 9.6Kbits/s over a standard GSM phone link, I can back up the data as I produce it. I can also retrieve the data from anywhere in the world - as long as I know the URL of my backup server.

This time next year, most notebooks will be sold with not just Wi-Fi built in, but also dual Wi-Fi and Bluetooth. Even infrared is making a comeback, as people come to understand it's sometimes an appropriate solution.

The new 50Mbit/s Wi-Fi5 standard will almost certainly be included in high-end systems as well. Intel has already demonstrated 802.11a alongside 802.11b in the UK and it's selling 802.11a systems in the US now.

Anyone who wants to know which standard to go with is missing the point. What's far more important is to know how to manage a corporate network so that users of all systems have the appropriate support.

Even to my inexpert eye, the flat-screen display on my colleague's desk was something special. Big, clear and responsive, it had that unmistakable 'you can't afford this' look to it.

It was Eizo's latest FlexScan L685, which is, unmistakeably, a top-of-the-line display. I asked him what it cost. "Ooh, this is the expensive one," he said. "How much?" We took bets.

He thought it was around the £3,000 mark; I thought it might be under £2,000. We checked with Eizo and found it was £1,109 (ex VAT).

When flat-screen displays first appeared, they were all unaffordable. These days, anybody with a cramped office will probably go for a smallish flat-screen display, rather than a larger CRT box, even if they feel they'd prefer a larger screen area, better colour balance or a faster response screen.

But when it comes to big displays, we feel that the colour LCD is simply going to be out of reach.

I think this display marks the point where the makers of cathode ray tubes have to stop and ask: "Do I want to stay in this business if it stops being a mass market?"

There is a demand, for sure, for CRT televisions. But, increasingly, it's become a widescreen market. Widescreen CRT isn't something that computer display technology uses comfortably. To put it politely, there are 'driver issues', which can be fixed, but why would you? A widescreen CRT is an expensive display, but you could get the same format, if you wanted, on a smaller LCD.

The world itself is widescreen. If you stand outside and look around, most of what you want to look at is on the horizon - the wider your vision, the better. But the world of data is actually more vertical. We use panoramic displays on computers because that's what the CRT makers were making for TV when we started using PCs 30 years ago. We've adapted to this, but the fact is that for editing text, a portrait-oriented display is far more usable.

Again, there are driver issues, but the one thing you really can do with a flat-screen display is have a dual-standard. Anybody who remembers the old pivot-rotating CRT will know that a huge glass box of vacuum is unwieldy and that if you have the thing set in one orientation, it's impossible to switch. Apart from anything else, people put toys on top of their CRTs.

If you ran into a driver issue when playing with a vertical-oriented display, it was a right royal pain to switch. But with a flat-screen display, it's just a moment's exasperation and a rotate and carry on.

It's too soon to say that portrait-oriented displays will be with us next year and it's probably too soon to say that the CRT is dead for PCs. But my money will be on the majority of mid-range PCs being bought with flat screens within two years - and possibly within 12 months.

When the BBC found it was able to buy electronic broadcast devices on the open market, rather than having to design them from basic research up to production, it did its best to keep its famous R&D department going in the old manner for as long as possible. In the end, however, the operation was heavily scaled down. After all, why do original research on transmitter technology just to see commercial organisations make the profits?

In a very similar way, the technology used to switch phone calls and internet traffic is no longer something a telco can manufacture for itself - or even specify. A mass-market box costs a fraction of a purpose-built one and British Telecom has had to face this fact. It too has a huge, resourceful and highly reputed R&D establishment, but also a serious problem: what to do to justify keeping the Martlesham Heath researchers together.

In some ways, the problem isn't going to go away. In others, the BT Adastral Park site (the new name) has a future where the BBC's research arm didn't: it not only owns 18,000-odd patents, but it has the money to turn them into businesses. And that's what it's now doing.

Of course, the majority of the national press comment about BT focuses on its huge debt. Indeed, this isn't fantasy; it does have huge debts and is struggling to pay them off out of its income. But there's a proportion issue to consider. The income might be small compared to the debts, but the amount of money needed to turn a few basic BT patents into a business is trivial compared to BT's income. And the potential rewards are absolutely huge - many of the patents are aimed at the IT business, not just at the telecoms industry.

On a recent trip to Adastral Park, I saw some fascinating - and apparently original - research into subjects like parasitic networks, where you do away with the need for fixed 'cells' to support mobile phones by setting up ad hoc mesh networks where individual devices instantly forward data to the nearest individual in the right direction.

I also saw a pure IT proposal turning into a business, APSolve, which is nothing more than a commercial exploitation of the software BT uses to direct its 22,000 vans to the 150,000 tasks they do each day.

There was also the Gabrielle voice portal, which is the clever systems integration of voice control with mouse control for Web portal access from a variety of information appliances. The basic technology is clever, but it's simple systems integration discipline that turns this into something to attract venture funding.

I suspect that most of these startups will fail. Out of every 50 ideas, BT Brightstar - the funder of the R&D spinoff project - reckons only one will become a business. And of those, one in four might become a corporation that can play globally with a market capitalisation of £50m and a long-term future.

It sounds like carnage, but in fact, with BT now funding over a dozen new companies a year, this could mean three new corporate technology-based giants rolling out of Suffolk every year in a very few years' time. I like to see centres of excellence and I like to see them succeed. I'm very much in favour of wishing the Martlesham experiment well.

I've been watching a lot of video on my notebook PC recently. Things like video chat, conferencing and Bloomberg TV. It's changed my requirements for processor design.

A year ago, a 'portable' PC would typically spend most of its time on the same desk. Today, that's just not acceptable; it has to be lightweight enough for you to carry around. Today we want smaller batteries and clever power-saving tricks. Unfortunately, though, these new applications mean that cache can't help us save power.

For someone who's just typing or making notes, cache algorithms can switch the disk off, slow the processor down to a trickle and turn a two-hour battery life into four hours - especially if you can switch off the backlight.

But increasingly, people want to watch streams of multimedia. And the research effort is going into bandwidth optimisation. However, bandwidth isn't the problem. The problem is that even small video streams - like the sort you get from conferencing - require quite a processor load for compression and decompression, as well as a lot of disk buffering. Buffering isn't the same as cacheing when you're dealing with real-time data.

If I want to watch a pre-recorded video stream, I can (in theory) download the load on a super-fast link, then play it back later from local storage, freeing the link. But I'm still grinding on with the hard disk and processor.

Intel has solved its speed problem with the Pentium 4. What I want to see in 2002 is some sign that it understands the power consumption problem.

Like this story? Spread the news by clicking below:

 del.icio.us     Digg this     reddit!