finally deal with flat variables structure

The robot is still based on a flat structure of variables. The lack of namespacs or the ability to limit access to variables only to the resourcefiles level has been mentioned for a decade in various issues, which continue to hang as suspended in limbo:

https://github.com/robotframework/robotframework/issues/1999, https://github.com/robotframework/robotframework/issues/4624, https://github.com/robotframework/robotframework/issues/2970

Parsers are undergoing another reorganization, and the mentioned problem seems to be overlooked.

The robot has developed to a level where it can quietly qualify as a programming language, it is used for more and more advanced projects, and sooner or later everyone will face the problem of lack of namespaces. As long as someone has experience, it's not a problem, because he will virtualize namespaces for himself using variable naming conventions.

Worse when one wants to introduce new people to the world of RobotFramework, or try to convince someone to use RF as an automation tool and put in the CONS section: no namespaces support for variables.

Unfortunately, I'm not proficient enough to test this myself, but is it impossible to use the same collision detection mechanism for variables coming from different resources/variablefiles/.etc, which catches collisions in keywords coming from different places?

Start by reading the linked issues 1999, 4624, and 2970, then review the current parser reorganization mentioned in the issue. Trace how variables from resources and variable files are represented and how keyword collisions are detected. Done means a defined namespace or resource-file access model with documented collision behavior and coverage for the affected cases.