“…This is because our definition relies on control of the trace norm of the non-self-adjoint operators a(1 + D 2 ) −s/2 , a ∈ A. It is shown in [CGRS2,Propositions 3.16,3.17] that for a spectral triple (A, H, D) to be finitely summable with spectral dimension p, it is necessary that A ⊂ B 1 (D, p) and this condition is almost sufficient as well. Anticipating the pseudodifferential calculus, we introduce subalgebras of B 1 (D, p) which 'see' smoothness as well as summability.…”