Currently, from the Boron release onwards, the YANG parser allows to provide a list of features that are supported, and it will transparently remove the unsupported nodes from the resulting SchemaContext.

This implies that you end up with a separate model for each device that is connecting (well, with sharing of similar models), but it also means that you can only know the correct model once the device has connected, because only then the supported features are known.

In a case of pre-provisioning you would like to have the correct model upfront. The idea is then that at the moment that you push that to the device, you remove all the nodes that are not supported in that specific device based on the if-feature information. However, with the current YANG parser support this is not possible, or you would have to make models with all possible combinations of features. We can create a model that contains all features (by saying that all features are supported), but the model then does not give information that some nodes are feature-dependent.

So we request that a separate mode is introduced where you don't provide the features in advance, and the SchemaContext then contains feature information per SchemaNode.