Full Idea
There are two difficulties with Carnap's taking possible worlds as linguistic. Everything must have a name, or our state-descriptions will be silent about nameless things, and nothing may have two names, or we may affirm and deny a predicate of one thing.
Gist of Idea
Linguistic possible worlds need a complete supply of unique names for each thing
Source
David Lewis (On the Plurality of Worlds [1986], 3.2)
Book Reference
Lewis,David: 'On the Plurality of Worlds' [Blackwell 2001], p.145
A Reaction
The idea of possible worlds as linguistic has no appeal for me, so this problem doesn't surprise or bother me, but it sounds fairly terminal for the project.