Companies crunch public stats for services to developers and administrators
by Andy Oram
Companies are constantly opening new veins of ore as they attempt to mine the Internet for useful information. Developers and open source system users will be particularly interested in a SourceLabs announcement of a service called Self-Support Suites that has been in beta since December. This tool combines enormous amounts of information indexed by SourceLabs from bug trackers, technical mailing lists, and other sites to help open source users diagnose problems. They've just put up a free download.
The proof of concept I heard from Byron Sebastian, CEO of SourceLabs, concerned a site that spent two weeks trying to track down the failure of an Apache Project module. SourceLabs's system found a bug report with the fix in a few minutes by finding a match between a stack trace provided by the user and a stack trace provided by a question in a public forum message. This search was more difficult than it might sound, because stack traces don't match precisely and their contents are not unique strings that are easy to search for. Sebastian says that stack traces and log files tend to have the most useful information--but if other information was organized better, it might rise in value.