Internet grinds to a costly halt

The information superhighway turned into the information bridleway yesterday, as the after-effects of a hacking incident last week caused the network in the United States to slow almost to a crawl.

The incident was the worst crash to affect the network since 1988, but affected far more people because of the Internet's growth since then.

Millions of Americans found that, as far as the computers which operate the network were concerned, they had ceased to exist. And although the problem had begun to be cleared up by midday yesterday, the repercussions may be lasting.

Yesterday's slowdown involved the essential "root servers" which automatically look up millions of Internet addresses every day and convert them into machine code. Anyone trying to access a machine whose Internet address ended in ".com" - the default for any American company - or ".net" - the default for companies offering Internet access - found their browsers and e-mail stalled.

Search engines, used to look for data amidst the gigantic World Wide Web, were particularly affected. Without being able to use them to search for data, much of the information on the Internet became as useless as a jumbled dictionary.

Blame at first fell on Eugene Kashpureff, the operator of an Internet company called Alternic, based in Bremerton, Washington state. Last week he altered the software on the root servers belonging to Network Solutions, a privately held company based in Boston, which allocates commercial Internet names (such as ""). When users tried to access Network Solutions' systems, they would be directed to Alternic's page.

But though Network Solutions said yesterday that it had solved that problem, British Internet companies said that in doing so, the company may have led to yesterday's slowdown. "I think that in rebooting their systems they screwed up," said Justin Keeney, technical director of Cerbernet, a UK Internet service provider.

The "root servers" are essential to the Internet. They perform the same process as a directory enquiries service, but are far more essential because there are far too many machines on the Internet for any single machine to be able to list them all.

Instead, when a user enters the name of a Web site - such as "" - another computer analyses the name, starting at the back. The root server would identify that address as being in the UK, and belonging to a company. It would then direct the enquiry to a "name server" dealing with UK companies.

But yesterday it was suggested that Network Solutions had created a bug in its own systems so that seven of its eight root servers were flawed. Only two other root servers, both belonging to the US military, were still working yesterday afternoon.

Mr Kashpureff and his company, Alternic, could not be contacted yesterday: the phone had been cut off. He, however, has been a long-standing advocate of an expansion of the Internet's naming scheme so that ".com" could be replaced by a virtual infinity of names such as ".newspapers", ".sport" and so on. The US recently forced through such a move, expanding the number of possible "domains", against the wishes of European companies. It may now decide that it was premature.