Novemberfest and Conventionalism

A busy couple of weeks at Arche, with two workshops (on semantic paradox and vagueness respectively), and talks by (among others) Greg Restall, JC Beall, Hartry Field, Diana Raffman, Achille Varzi and Dominic Hyde (with many others in town - e.g. Steve Yablo, Richard Heck, Graham Priest). I'll try and post some paper reports here as the fortnight progresses.

In the meantime, does anyone think the following is a way to rescue conventionalism about necessary/a priori/analytic truth from one obvious type of objection to it?

Here's the objection (as expressed by BonJour):

[w]hat convention might be adopted that would make it possible for something to be red and green all over at the same time? It is, of course, obvious that new conventions could change the meaning ... of the words ‘red’ and ‘green’, but there is no plausibility at all to the idea that such changes would result in the falsity of ... the proposition that nothing can be red and green all over at the same time, as opposed to merely altering the way in which [that proposition is] expressed.
(From In Defence of Pure Reason, p. 53.)

Mightn’t the conventionalist try and distinguish two ways of understanding the claim that had our conventions been different it would have been possible for something to be red and green all over? On one of these, the relevant counterfactual worlds are being assessed by us, and therefore our own conventions are in play. So we deny that these worlds where our conventions are different are worlds and something can be red and green all over (because our actual-world conventions fix that nothing can be red and green all over in any world, including these ones). On the other approach we consider, not what is true at those worlds considered as worlds governed by our actual conventions, but what is true at those worlds considered as worlds governed by the conventions we have at the worlds in question. So on this second approach we accept that there are some worlds where it is possible for something to be red and green all over at the same time.

Maybe the conventionalist could argue that our intuition that changing our conventions wouldn’t change the facts (the intuition driving BonJour's objection) is well-enough preserved by the result we get on the first approach. But on the second approach there are worlds where the proposition is made false by the fact that we have different conventions at those worlds. And this (she might say) is enough to rescue the thought that we could have had different conventions which would have made it false that nothing is red and green all over - i.e. enough to rescue conventionalism from the objection.