Human error, that ever-unpredicatable factor, brought the internet to a virtual standstill yesterday in many places when the InterNIC database – effectively the internet’s address book – was wiped clean by its keeper, Network Solutions Inc (NSI). Anything on the internet with a .com or .net address was affected to some degree and meant that e-mail could not be delivered and websites were not accessible. A Computer Associates International Inc Ingres database at NSI crashed, corrupting the zone files that hold the domain name information. The crash triggered NSI’s own quality assurance software to kick in and alert the adminstrator to the problem, but the administrator ignored the warnings and sent the files out to the internet anyway. NSI claimed the problem had been rectified by 6:30 Eastern Time Thursday morning. The company declined to comment on what action was taken aganinst the administrator, other than it was appropriate. But that was obviously too late for Europe. While most of the US was still in bed, e-mail that Europeans sent to these two network domain names was being returned because it did not know where to be delivered, or bounced back, as it’s called in netspeak. NSI currently runs the internet’s domain name system, athough there’s a movement afoot to open the name registration process up to other organizations. The corruption meant that most users’ domain servers were unable to locate any domains with the two most popular extensions. Regularly-hit websites’ domain names are cached – or put into local storage – on the users’ servers, but internet addresses not cached have to be looked up in the InterNIC database by a process known as a domain name server look-up. That’s where domain name servers come in. They match the names with IP internet protocol numbers that should enable users to locate domain servers anywhere on the internet. There are eight of these servers in the US, and one each in London, UK and Stockholm, Sweden. A Japanese server is planned. Not all the servers were affected; Sweden’s escaped the problem supposedly because it uses some specialized Unix technologies. The accident happened when the servers were updated with new names, which happens at least once a day, though nobody was able to say exactly how many times. The update contained no information about the two top-level domains, meaning all the previous data was overwritten with blank entries. The servers had to wait until the next update to receive data about new domain names as well as the old data from a back-up. Herndon, Virginia-based NSI has the exclusive contract to manage the .com, .org and .net top-level domain names. It is supposed to administer the InterNIC jointly with AT&T Co and the National Science Foundation (NSF), though in practice the InterNIC is NSI. The NSF was unaware of the problem when we called yesterday. NSI was granted a five-year contract by the NSF in April 1993, and is confident that it can carry on registering and maintaining the names after the contract expires in March. But the NSF says it will neither renew the contract nor give such monopolistic control to any other company. Last month, NSI filed for an initial public offering, attempting to cash in on its position while the going was good. That filing raised eyebrows at the US Department of Justice, which launched an anti-trust investigation into the company.