Allowing customization of telemetry through user agent

[liu-ziyang]

What's the problem this feature will solve?

We are developing a desktop application that uses pip to install plugins, and we want to understand the installation metrics of how people are using the application. To do that we want to customize the telemetry that pip sends, which is later aggregated by the linehaul project to BigQuery.

Describe the solution you'd like

Having an optional flag or argument that allows overriding of the telemetry user agent that pip used to send installation reports

Alternative Solutions

We attempted to use the PIP_USER_AGENT_USER_DATA. However, the user_data is currently ignored by the linehaul parser and never made to the telemetry dataset

Additional context

Original discussion on linehaul-cloud-function:
pypi/linehaul-cloud-function#94

Code of Conduct

• I agree to follow the PSF Code of Conduct.

[uranusjr]

I’m inclined to say this is not a general enough use case; this should be considered redistribution and providing custom patches is more appropriate. We could make the code easier to patch, but an environment variable feels too much to me.

[pfmoore]

Agreed. We have PIP_USER_AGENT_USER_DATA for this purpose, and if linehaul is ignoring it, then that's either a linehaul issue, or you are using the data collected by linehaul for a purpose it wasn't intended to cover. (My feeling from the discussion on the linehaul issue is that it's the latter).

I will note that your application could run pip via a proxy (which could also be running in the application, so no external software needed) which modified the user-agent header in whatever way you wanted. That would avoid any need for changes to pip.

[liu-ziyang]

Thank you both for your quick triaging and response! I am going to close this request now and maybe attempt to integrate with proxy later as suggested