Cache Flow Problems

Last Friday morning, Andrew I. Jones — a McKinleyville-based publicist, Web designer and publisher — logged onto his popular parenting blog, Thingamababy, to add a comment to one of his posts. But he was having some problems. He refreshed and refreshed, but thingamababy.com wasn’t showing his comment. So he tried again, but received the same result. The administrative portion of his site showed that he had posted the comment twice, but the public part of the site remained unchanged. The comment was nowhere to be seen.

Meanwhile, over on the popular Humboldt Herald political blog, some people had been noticing odd Internet behavior since the night before — odd, and arguably somewhat sinister. Like Jones, readers of the Herald noticed that pages on the site weren’t changing very often. New posts and new comments were unavailable to some users. Even more strangely, the comment forms on the site — the place where people can enter their thoughts on the topic of the day — came pre-filled with the names or pseudonyms of other readers, as well as their e-mail addresses. In a forum where anonymity is often at a premium, this appeared to some to be a serious and frightening breach of privacy. And it wasn’t just happening on the Herald. Other local blogs, such as Redway attorney Eric Kirk’s SoHum Parlance and the anonymous Humboldt Mirror, were affected as well.

Readers of the Herald compared notes and quickly came to a conclusion. The users being affected by these glitches in the Internet all had one thing in common: They were all customers of Suddenlink, the cable company/Internet provider that serves a large portion of Humboldt County. They reported that sites all over the Web, including the North Coast Journal‘s, were affected to one degree or another. Problems mounted. Readers were confronted with pop-up boxes asking if they wished to “download” a Web page, rather than merely visit it. Visitors to the Herald were, at times, given a version of the site optimized for mobile phones, despite the fact that they were sitting at their home computer.

Jones reported the matter to Suddenlink Saturday morning, but received no response from the company for a couple of days. In the meanwhile, an impromptu mailing list amongst a few local technical professionals and website publishers sprung up, and this list quickly came up with a provisional theory: The cable company appeared to be caching Web pages — saving them on their own machines, then serving those saved pages to their customers. Every so often, it appeared, the company would refresh the cache with updated versions of the pages. This is not a completely unknown practice, but something was terribly wrong with the way Suddenlink was going about it.

In the case of Wordpress — a popular publishing platform used by all of the blogs mentioned above — it appeared to work like this: One user would post a comment with his or her name and e-mail address. The Wordpress system would send that user back the same page, updated with the new comment and with the submitted name and e-mail addresses entered into the new form. Suddenlink, it seemed, would cache that page, and then serve it to all other customers — name and e-mail included.

Perhaps no one saw the danger quite as quickly as Mitch Trachtenberg, a computer programmer in Trinidad who wrote the software used in the Humboldt Transparency Project, an open system for verifying the results of local elections. On the Humboldt Herald, Trachtenberg’s name and e-mail address were among the first to be handed out to other users of the site who happened to be on Suddenlink. He wasn’t too concerned with the breach of privacy in his own case, but he quickly saw how such a thing might affect others who might wish to keep their identities or contact information secret.

“If I’d cared about anonymity, I would have stopped using my real e-mail address when that started,” Trachtenberg wrote the Journal. “I haven’t been harmed, except by lousy service. I think others may have been harmed severely.”

By Monday, everything seemed to have returned to normal.

Reached Tuesday at his office in St. Louis, Mo., Suddenlink Senior Vice President of Corporate Communications Pete Abel knew that a cache system had recently been installed in Humboldt County, but was unaware of the particular problems reported by users. After speaking with the Journal and other Suddenlink employees, though, he released a statement explaining what appeared to have happened.

According to the release, the cache system was installed in Humboldt County on Thursday, Jan. 27 — the very day that users began experiencing problems — and was intended as an interim solution to relatively low Internet speeds in Humboldt County. The system, it said, was able to cache only unsecure websites — those which, unlike almost all reputable banking or commerce systems — do not encrypt communications. But the company eventually discovered the problems that its customers had been reporting and, having fruitlessly worked with its vendor to find a solution, turned the system off on Monday.

“The good news is that secure Web site pages will not have been cached,” Abel said in a follow-up call to the Journal. “And I have been assured 100 ways from Sunday that never would have happened.”

At the same time, though, a great deal of login information for many noncommercial Web sites is non-encrypted. Though no one has reported even greater breaches of privacy than the ones experienced by Trachtenberg — which also affected other people in the county — it is not impossible that they may, in fact, have taken place.

Jones, for his part, was far from pleased with how Suddenlink technical support handled the issue. It took two days for the company’s technical support staff to get back to his request to opt out of the company’s caching system, and when it did it informed him only that he would have to upgrade to a business account if he wanted that service. “The residential service is for entertainment only,” a company representative wrote.

For Jones, that answer isn’t good enough.

“If a small local radio station intermittently went off air for multiple days, the radio host would be apologizing and explaining the situation,” Jones wrote the Journal. “If a large utility company experienced sporadic power outages, people could hear a recording on a toll-free number to learn the cause and about ongoing repairs. What does an Internet provider do when web access becomes spotty and begins serving customers old copies of web pages? The company gets back to you in a couple days and suggests you pay more if you don’t like its recently degraded services.”

This article appears in Point-In-Time.