This question may sound silly, but there is something about the semantics in first order languages which confuses me a bit.
Im using Ebbinghaus,Flum,Thomas:Einführung in die mathematische Logik, as well as Shoenfield:Mathematical Logic for reference
They define (I translated the :gdw to :iff):
$$\vDash \mathfrak{I}(x_1 \equiv x_2) \qquad \mathrm{:iff} \qquad \mathfrak{I}(x_1) = \mathfrak{I}(x_2)$$
and for Shoenfield, he uses no semantical symbol but introduces the term structure and then says for the formula $\mathbf{A} \ which \ denotes \ \mathbf{a} = \mathbf{b}$
$$\mathfrak{S}(\mathbf{A}) = \top \quad \mathrm{iff}\quad \mathfrak{S}(\mathbf{a}) =\mathfrak{S}(\mathbf{b})$$
Which seem very similar (presumably they are actually both describing satisfiability relation and/or validity relation), but for a few things.
my question: I understand how I can use "satisfiable under an interpretation" to eliminate junctors by using the truth function for those, however with these definitions (similar holds for a general relation) I still technically dont see how I would be able to formally be able to say that $a=b$. For while Ebbinghaus states a rule (which actually seems like an axiom to me) that reflexity of the equality can be infered without an antescenden, and Shoenfield has equality axioms, either are syntactical (right?) and thus unrelated to the evaluating the equality under an interpretation. What am I missing, or is it implied that the use of the interpretation is done informally (unless the meta language is formalized as well)?
I will bring an example similar to one Ebbinghaus brings to make it clearer:
Let $S = (+,\cdot,0,1)$ on$\mathbb{N}$ be a structure, and $\beta(v_n) = 2n$, then the formula $$v_2(v_1+v_2) \equiv v_4$$ under this interpretation would become $4 \cdot(2+4) = 8$ (note they use $\equiv$ before interpreting and $=$ after, so I presumed that the evaluation is actually done in the metalanguage, while the equality axiom is stated in the object language etc