Open jrieken opened 2 years ago
Yeah this is a bit more of a complicated refactor which is why we steered away from it. The render processor implements the common telemetry service which is implemented pretty much everywhere telemetry is sent. This is required because some things like workbench common properties are unique to the renderer. The "sending" in the shared process is just all the data already formed and cleaned being sent over IPC and then directly to app insights.
Yeah, I understand that's nothing for November but generally this should be relatively easy, right? Cleaning is done with a bunch of regular expression and if those can only be created with the renderer help it should be easy to send them over to the shared process, right?
Follow up from https://github.com/microsoft/vscode/issues/138153
We do a bunch of cleansing operations so that we lower the chances to accidentally leak PII or any other sensible data to telemetry. This currently happens in the renderer process and therefore can impact performance, esp. startup performance. I also don't know how/if we clean telemetry from other processes, like the main process. Since telemetry is send from the shared process we should do the cleansing there so that even mistakes don't impact startup performance