The paper tackles a question in computable mathematics: which function spaces, specifically the classical Lp spaces used throughout analysis and probability, can be given a "computable presentation"? A computable presentation means you can represent the space in a way that a computer could, in principle, work with the elements and operations effectively. The spaces in question, Lp[0,1] and their sequence-space cousins lp, are families of spaces indexed by a number p greater than 1, and the paper asks how the computability of the space relates to the arithmetic nature of the parameter p.
The main results give a surprisingly clean picture. For the continuous function spaces Lp[0,1], a computable presentation exists as long as p is a "right-computably-enumerable" real number in the range between 1 and 2, a condition slightly weaker than p being fully computable. For the sequence spaces lp, the situation is stricter: having a computable presentation forces p itself to be fully computable. Together with previously known results covering other ranges of p, this completes a full classification of which separable Lebesgue spaces admit computable presentations. The contrast between the two cases is notable and reflects a genuine structural difference between the continuous and sequence-space settings.
The technical machinery developed along the way is independently interesting. The authors build an effective version of a classical construction called a Fraisse limit, which is a way of assembling a large well-behaved mathematical object by carefully piecing together smaller ones, and they make this process algorithmic for Banach spaces. As an application beyond the Lp world, they use this framework to give a computable presentation of the Gurarij space, a special Banach space that plays a role analogous to a universal homogeneous structure, similar to how the rational numbers sit inside the reals.