On Thu, Oct 23, 2014 at 8:51 AM, Aaron Heady (BING AVAILABILITY) < [email protected]> wrote:
> We’ve been pretty careful to this point about including very little (no) > user sensitive information in the error events. Although I would love to > get the stack trace, I think that would require something like opting into > a ‘Customer Experience Program’ that has informed consent. Stack traces > from apps contain a lot of potentially sensitive information. Much more > than the URL and other normal request information that we technically > already have access to when a UA makes a request. > +1. Practically speaking, I think stack traces are off-limits. > I was thinking about it just being a specific error type, say ‘Crash’ > versus ‘DNS failure’, and we’d get the same information as other errors. > The key being the URL that the crash occurred on. > We have "abandoned", which maps to user canceling the navigation. Perhaps adding an error type that indicates a "UA abandoned" state is not unreasonable? This could be due to crashes and other processing errors - e.g. aborted due to a redirect chain. ig
