Figure 1-B: clearly indicates that first the expression is presented in CQL and then it is parsed into the syntax tree. Therefore the text here seems to be missleading.

Moreover, even if the representation is as an abstract tree, unless the representation assumes binary object dumping, applications can not avoid parsing these representations. If the ELM is provided to an application in a text format within a file then the application still has to parse the tree to, for instance, create the object model in memory.

So please either re-work the justification, by clarifying the assumptions, or remove it. From the application stand point, there is no fundamental different in whether to parse the expression or the same expression serialized as a tree. The only difference is if the tree is already present in memory in a kind of object-model form that the application can directly work with in an object oriented faschion.

Existing Wording:

As shown here, the ELM representation is defined as an Abstract Syntax Tree, eliminating the need for lexical analysis and parsing steps, and allowing implementations to concentrate on the core representation of the logic.