potiuk commented on issue #38745: URL: https://github.com/apache/airflow/issues/38745#issuecomment-2039774537
> In my opinion, the main problem with having templates being rendered in this case is that now we need to sanitize user data when processing it in expanded tasks (I didn't think of that at the beginning, but it is probably also a security issue in this case). When using plain tasks you generally don't need to be concerned whether the data you are processing can be harmful outside of the user code inside decorated python function context. Personally I think that would be unfortunate if this becomes a default behaviour for expand as well I don't think there are any significant security implications here. The only code that can really get executed and templates that are coming in the DAGs come from DAG Authors. There is no code that can be executed that can be supplied (as far as I know) by someone else (i.e. UI users - except those users that have "Connection Editing" capabilities. And this is all within Airflow Security Model: https://airflow.apache.org/docs/apache-airflow/stable/security/security_model.html - DAG authors already have capabilties to execute anything they want and there is no way nor intention to block them from doing so (this can be controlled and verified at code review / static checks on DAGs outside of Airflow). Unless there is a scenario that other airflow users can provide a code that can be executed in the context of workers, I do not see how template expansion could go outside of the security model we already have and know. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
