Just over a week ago, while searching for a file on a Windows-XP machine, I was surprised to see the Search Assistant attempting to activate my Internet connection. It puzzled me because I wasn't searching the Internet, only my local drive. I was busy with other things at the time, but I made a mental note to look into it soon, which I promptly forgot to do.
This morning, Reg reader Jody Melbourne rattled my cage, fresh from having made the same discovery. He'd noticed that the Assistant was establishing a connection with a machine at Microsoft.
"I did not give Microsoft permission to know what files I am searching for on my local hard-drive," Jody wrote.
Indeed, and neither had I. So I connected an XP box to my ISP, started a packet sniffer, and launched the Search Assistant. Sure enough, it immediately connected to http://sa.windows.com/ and fetched a number of files. But it didn't attempt to send any data to the site, beyond comparing my locally-stored versions of those files to the ones on the server.
But when I performed an Internet search, the Assistant sent my search terms to the Microsoft site, and also dropped a session cookie on my machine.
One of the files the Assistant fetches is the MS Search Companion privacy statement. This is done for P3P compliance. According to the statement, MS doesn't collect information about local searches. "No information is ever collected by Search Companion when you search your local system, LAN, or intranet for any reason."
I certainly didn't pick up anything to contradict that. But there is some obvious collecting when SA is used to search the Internet.
"When you search the Internet using the Search Companion, the following information is collected regarding your use of the service: your IP address, the text of your Internet search query, grammatical information about the query, the list of tasks which the Search Companion Web service recommends, and any tasks you select from the recommendation list."
"Search Companion does not record your choice of Internet search engine, and does not collect or request any personal or demographic information. Information collected by the Search Companion cannot be used to identify you individually, and is never used in conjunction with other data sources that may contain personal data."
Hopefully there aren't too many loopholes in that, though I rather think the user's IP can be considered personally identifying. However, MS tells us that the policy statement is out of date. IPs were logged for testing purposes during the XP beta period; but since the product launch, there has been no IP logging.
In addition to the privacy statement, the remaining files fetched are XSL (Extensible Stylesheet Language) stylesheets:
Users curious to know exactly what they contain can quite easily locate them on their local machine and have a peek. According to MS, they're simply used to maintain up-to-date associations between file extensions and file types, to make searching more productive.
I'm not acquainted with XSL, so I'm in no position to affirm that or to argue with it, but I'd be pleased to hear from readers who can shed additional light on the subject.
For now it appears that there's nothing here for users to worry about. But there is a question about MS playing fast and loose with people's Internet connections. Certainly, the minute one ventures onto the Web, one starts bleeding information all over the place, fetching images and ads and taking cookies from secondary and tertiary sources too numerous to mention.
But when we run an application for some local business like a file search, we don't expect it to connect silently to the Net, even for a good reason. When we discover something like this, it feels like someone else is in control of our computer, and that is definitely not a good feeling.
If Trustworthy Computing is going to mean anything, it's going to have to mean that actions like file downloads aren't going to happen without the user's knowledge and consent. A simple popup asking if one wants the latest XSL files with the options to decline, to be asked each time, or to grant permission to go ahead without further consultation is all that would be needed. ®