A telescope is the variable context in a dependently typed program. Dependent types with free varialbes are really only well-defined in a telescope. A segment of a telescope is a tuple of types in that telescope. These are usually implicit in formal developments, but seem to be crucial to get a compositional account. Here we use them to get 'n-ary dependent function' (compare with `Data.Fun` and `Data.Rel`). I've learned this from Conor McBride on an SPLV'19 bus ride. A literary reference would be welcome. This paper is a start: Unifiers as Equivalences: Proof-Relevant Unification of Dependently Typed Data. Jesper Cockx, Dominique Devriese, Frank Piessens, ICFP'16. but they don't seem to have segments and their left action on contexts.