But because the Internet is so vast and complex, it is impossible to know for sure exactly why surfers were unable to reach some Web sites at certain times yesterday morning, said David Holtzman, senior vice president of engineering for Network Solutions, the company that holds the government contract to run the registry for the most popular top-level domains--".com," ".net," ".org," and ".edu."
There is disagreement on what caused the problems. Whether or not yesterday's glitches caused the reported access problems remains unclear. But starting early yesterday morning, Holtzman said Network Solutions noticed a problem with one of the 13 root servers that host the domain-name system.
Several others in the Internet community, such as system administrators, said the problem was caused by faulty information generated by Network Solutions' own servers.
The servers essentially act as the Net's telephone directory. Every day, they receive the addresses of all domain names from Network Solutions. Internet service providers in turn get information from the servers and provide that information to the end user.
While the system is invisible to the end user, without it, Netizens cannot reach their destinations in cyberspace. In other words, the domain name system acts like the a telephone operator, switching calls to the correct place when someone punches in a certain number.
The 13 name servers are run by separate entities. The people and institutions that run them work largely on a volunteer basis, and always have worked by consensus.
When one of the servers fails, ideally, the others pick up the slack. While it remains unclear whether such a failure is what caused yesterday's glitches, NSI's Holtzman said that the F root server (the servers are named alphabetically) "was returning erroneous information."
Paul Vixie, whose Redwood City, California-based Vixie Enterprises manages the F server, could not be reached directly for comment, but last night he posted a message to The North American Network Operators' Group mailing list saying that the problem had to do with problems at Network Solutions.
"F root-servers.net was lame for two days because it could not fetch [".com"] from NSI," he wrote. "*MY* server was not sending fatally bad answers, *THEIRS* were."
Several other system administrators also said that the problem was caused by NSI and not by Vixie's machine.
Holtzman and another NSI spokesman denied yesterday that the problems were caused by NSI or that they went beyond the F root server.
Small problems in the root servers, Holtzman added, are "routine," and usually go unnoticed. "On any given day, we have some kind of problem with one of the root servers. It's fairly typical. This is a little worse in that it's returning something, rather than nothing."
"The Internet is a big organism," he said.