The fatal accident looming on the old superhighway—if they dont find something magic to do about it—will be the 3G (third-generation) phone networks.
At the recent UMTS show in London, the presentation that caught my attention was a "speed up wireless" plea from Yair Shapira of Flash Networks, a presentation that basically said: "Your 3G business model is broken."
Of course, he didnt put it quite as I did in the opening paragraph of this column. What he was actually doing was offering a solution to the horrible, unacceptable fact that an awful lot of perfectly ordinary data processing operations dont work on GPRS (General Packet Radio Service)—on 2.5G wireless.
And the solution, it turns out, is not to move to 3G, because its no better.
Shapiras illustration was the multiple handshakes and transactions between Web server and terminal.
On a GPRS link, it can take half a minute to download a fairly ordinary Web page. This is usually put down by the impatient user to congested networks (quite accurately, in most cases) or to feeble processors in the phone.
In fact, as Shapira pointed out, the underlying problem is the huge latency in the network and the number of items that have to be fetched. "The horrible truth is something Ive seen in laboratory reports from inside a big European 3G operator," he said. "It is that WAP2 over 3G is not better than WAP1 over GPRS. This big operator invested hugely in 3G in the hope of making WAP [Wireless Application Protocol] go faster; but in the lab, its not faster."
The problem is that the round-trip time of a 3G message and response is closer to a second than a fraction of a second. Shapira estimates that the typical time lapse between asking a Web server for a data item, such as a logo, and being able to render it on-screen is about 800 milliseconds.
Nobody whos used 3G data would contradict him, although some might say: "Oh, its not quite that bad!" and then settle for 400 milliseconds.
But a Web page may have a dozen to two dozen items on it. Each is fetched separately by HTML browsers. On broadband, with latencies in the 50 to 100 millisecond range, the cumulative delay isnt crippling. On 3G, a delay of two seconds turns, as if by having a spell cast on it by a wicked fairy, into a half-minute pause.
OK, so much of this is invisible to users, because users cant see the whole page. The screen is too small. So, they scroll painfully down, and the page carries on rendering in the background while they press the scroll button ... but the world is changing, and everybody knows it.
In a year, maybe, and definitely in two, this sort of performance will be starkly apparent and definitely unacceptable.