By doing this, we can also make sub-nodes for the "constant/literal" versions of IDs (the ones wrapped in quotes) vs the variable/non-constant ones (not wrapped in quotes).

At the moment, these cases are distinguished by checking the text in the node, which is not as good as having distinct types of nodes.

This is an issue that should probably only be worked on after we have done a language re-design on paper (because it may not be relevant afterwards).