Security warning too quick for comfort?

Network protection company Internet Security Systems published a security advisory for Apache, the Internet's most popular Web server, and gave the Apache Foundation, which created the software, less than two hours to respond.

Considering the potential seriousness of the issue, the company's public announcement of the problem without first talking to the Apache developers wasn't responsible, said Mark Cox, a founding member of the Apache Foundation.

"There are many minds on how long to give a vendor to respond," Cox said. "Some say until the vendor releases a patch, others say 10 hours. In any event, two hours is not sufficient."

Both ISS and the Apache Foundation released information Monday about a vulnerability that could allow vandals--using a simple HTTP command--to execute a so-called denial-of-service attack on the majority of the 10.4 million computers running the popular Apache Web server software. On far fewer of the installations, including older Windows-based versions and some newer 64-bit Unix-based versions, the flaw is more serious and could allow the attacker to take control of the server via the Web.

The warning's release reopened a long-simmering debate over how much time a security researcher should give a software maker to verify and fix vulnerabilities that could affect large numbers of computer users.

Chris Rouland, director of ISS's research and development team, known as X-Force, maintains that the company did the right thing when it released an advisory on the issue and included a patch as well.

"We are competing with the 10 million hackers out there, who are trying to break in to Web servers," he said. "The hackers were the real ones that were ticked off that we released the advisory. That's one less exploit that they could use."

Rouland said that by releasing the patch for the vulnerability, which ISS believed was limited to the Windows version of Apache, the company was being responsible.

However, Apache had already been working with another security researcher on the issue. Mark Litchfield, co-founder of Next-Generation Security Software, had found the flaw and told the Apache Foundation. Along with Litchfield, the foundation found that the flaw affects all versions of Apache, but to differing degrees.

"One of the problems with ISS releasing their advisory early is that we were trying to find out how different platforms were affected," said Apache's Cox. "We have lost that ability now."

He added that if the security company had contacted the Apache Foundation or the Computer Emergency Response Team (CERT) Coordination Center--a clearinghouse for information about vulnerabilities--it would have learned that the open-source group had already been working on the problem.

Because of the release by ISS, Apache had been forced into publishing its own advisory on Monday as well, Cox said. More than 70 software companies that distribute the software in various forms will have to play catch-up and create a patch quickly.

"With the premature release from ISS, many are now left vulnerable without a patch from the Apache supplier," security researcher David Litchfield, Mark Litchfield's brother and a co-founder of NGSSoftware, wrote in an e-mail posted to the popular Bugtraq security list.

Both ISS and Mark Litchfield decided to study the possibility of a hole after a similar problem known as a "chunked encoding overrun" was found in early April in Microsoft's flagship Internet Information Server (IIS) Web server software.

The reason ISS did not immediately contact Apache to notify it of the problem was because many members of the Web server project also worked for companies that compete with ISS or its clients, Rouland said.

"One of the challenges we are facing now in working with the open-source community is that no one 'works' for the Apache Foundation," he said. Instead, people such as Cox, who works for Linux leader Red Hat, donate their time to do the work.

"It may not be in my customers' best interest to let Red Hat know there is a security vulnerability," Rouland said. "I don't consider Red Hat a trusted third party."

The distrust puts open-source projects, and the users of their software, in a quandary if companies won't give them advance notice of security vulnerabilities. While some in the industry have suggested establishing a new vulnerability coordination center and others have promoted CERT as filling just such a need, it seems unlikely that the debate will get resolved anytime soon.