Full Idea
Class Nominalism can be defended (by Quinton) against the problem of random sets (with nothing in common), by giving an account of properties in terms of 'natural' classes, where 'natural' comes in degrees, but is fundamental and unanalysable.
Gist of Idea
'Class Nominalism' may explain properties if we stick to 'natural' sets, and ignore random ones
Source
David M. Armstrong (Universals [1995], p.503)
Book Reference
'A Companion to Metaphysics', ed/tr. Kim,Jaegwon/Sosa,Ernest [Blackwell 1995], p.503
A Reaction
This still seems to beg the question, because you still have to decide whether two things have anything 'naturally' in common before you assign them to a set.