Guy Kewney

I have met some people who live in fantasy worlds, but those who dream of turning GPRS into a serious proposition for computer users have to be up there with the wackiest of them.

As a frequent user of what is wittily called '2.5G phone services' on my trusty Thinkpad, I knew that all was not well with the world, but I wasn't altogether clear exactly why. Then someone gave me some statistics.

"You may be interested to know that there is an operator in the States that has a single 64Kbits/sec link to provide backhaul data for about one-third of the country," said one reader.

Provision for mobile data in Europe is substantially higher, of course. Exactly how much higher is a pretty tricky metric to uncover.

One of the problems that faces mobile data users - and not just over GPRS - is that the IT world talks a good fight about 'convergence'. But when you actually get down to it, there's a lot of giveaway signs proving that they aren't actually all that interested.

For example, when I was chatting to some Novell people at Brainshare a few months ago, one of the top technical guys was telling me about a particular new service the company was rolling out.

The details were impressive, the strategy inspiring, and it relied completely and absolutely on a 100Mbits/sec Ethernet connection. Anything slower, or with longer latency, and it simply doesn't work.

A consultant I know had managed - through extreme cleverness - to provide remote control services for his clients from anywhere.

What he'd done was obvious enough: he'd started using a virtual private network (VPN) from his notebook computer. He could dial into his ISP from anywhere, set up the VPN link to his client, and then take over their machine.

Such direct control eliminates all those tedious 'move the mouse up' instructions where the hapless user picks the mouse off the desk and holds it a foot above the desk. He could see what was happening.

Being clever, he one day decided to run his VPN across GPRS using his phone as a modem. It worked! It required substantial patience, but it eventually worked. And then a few months later, it stopped working.

I checked with the mobile networks. The only one that actually knew what was going on was Orange, and it knew because it was setting up a VPN-friendly GPRS service.

In doing so, it had discovered that most VPN software can tolerate IP address translation, but simply dies if you try to do port translation.

Orange now provides a genuine commercial 'VPN over GPRS' service. "It's really not easy, because none of the protocols are designed for this," said one of the technical staff setting it up.

"Almost all these commercial programs assume a latency of well under 100 milliseconds. Over GPRS, it is routinely three seconds.

"If you switch from one cell to another, you can have eight seconds' delay before data flow resumes and, believe it or not, you can switch cells without moving."

The thing that puts all this into perspective is the small number of people who use this 2.5G wireless for anything serious.

The amount of data that goes through a simple mobile phone looking at Wap sites is very small, hence the absurd charges per megabyte that GPRS fetches.

But plug a notebook PC into the system, and you're looking at spending your month's data allowance in one Outlook email session.

And the software, of course, isn't grateful; it chokes on the data retransmission delays faster than you choke on the bill.

So what are the phone providers doing? Actually, very little. What is happening, at last, is that they are really and truly starting to build third-generation networks.

In the UK this time next year you will genuinely have a choice of three 3G providers, including 3, in most urban areas.

So there's not that much incentive (or budget) for spending huge sums on building new data-carrying facilities onto the current GSM network.

What's worse is that the providers don't see the demand. They look at new IT-centre software and they see code which can't work at all over GPRS.

The typical FTP program, for example, is simply not capable of adjusting to the demands of a user who is travelling; and if you ask the authors to produce a new protocol, they say they'll doubt that there's any demand.

I'd like to be able to prove them wrong, but so far, all the evidence is on the side of the cynics.

In the circumstances, I'm starting to doubt the sanity of people who say: 'Data services and convergence are the solution to getting average revenue per user up and making phone networks profitable.'

At the rates they charge, nobody is going to shift much data. At the data amounts people will shift, there's no demand for infrastructure provision.

With the infrastructure we have, there's no way of boosting the amount of data we send to the point where it would make a serious market opportunity. Round we go, again ...

It's all, frankly, a chimera. A good joke for some phone users, but not to be taken seriously by computer users.

Like this story? Spread the news by clicking below:

 del.icio.us     Digg this     reddit!