[3.30.0-alpha] tags=nouserauth & [engine] userauth: registering for a new anonymous credential

[NewEraCracker]

Hi,

As the title says in relation of: #1803

The log entry:
[engine] userauth: registering for a new anonymous credential

Is registered for test operations such as:
ooniprobe.exe run websites

On the custom built with:
go build -tags=nouserauth -v -ldflags "-s -w" ./cmd/ooniprobe

Workaround:
ooniprobe.exe run --no-creds websites

If I disable the userauth binding, registering for a new anonymous credential should not be done (I think !?), as it is spamming the server with anonymous credential creation requests that will never be redeemed.

My two cents. Thank you.

---

Edit: This appears to happen only on a clean OONI_HOME.

I was able to build a probe binary with the userauth backend after fussing a bit more, just a ton of warnings during the build process. The issue may warrant investigation if desired, so I leave the decision to those that actually know the project.

[DecFox]

Hello @NewEraCracker, this is indeed a misleading log. Thanks for flagging this. The issue is we always try to submit with the credentials unless the --no-creds flag is passed in. No actual network calls are made if the userauth is missing since calls fallback to the stub, but the spam log comes from the caller. Let me see if we have a workaround that's less annoying.

Edit: This appears to happen only on a clean OONI_HOME.

Background: registration only happens if the user does not have credentials. A clean OONI_HOME, gets rid of user credentials. Source: https://github.com/ooni/probe-cli/blob/master/internal/userauth/submitter.go#L102, which is why this only appears when we have a clean cache. Since the probe never registers a credential with the nouserauth tag, it tries to register each time, which leads to a spam.

[NewEraCracker]

I am happy with your answer. This only happens once per run.

I'm currently focusing on detecting short to medium-lived repetitive content restrictions. Given the sensitivity of this topic, I'm keeping discussion limited to prevent unnecessary public visibility.

As I run the CLI in short-lived self-destructing docker-like environments, I currently do not see the need to integrate the userauth on my custom scripts.

The country was not mentioned on the report about tampered measurements, and the reason for anonymous credentials, so I currently feel comfortable with this approach. I may review the architecture in the future, and I'll sure be in touch privately if the need arises.

Thank you.