I wonder if the exact consistency strength of "All projective sets have the Ramsey property" is still open. In Solovay's model, all sets have the Ramsey property, so the consistency strength of this is below an inaccessible. As far as I know, there are some implications under forcing axioms. And all sets up to a certain level of the projective hierarchy having the Ramsey property is still doable from just ZFC. But I can't seem to find anything which shows Solovays inaccessible is really necessary in case of the Ramsey property (like Raisonnier/Shelahs result for Lebesgue measure), or oppositely, that you can get a model where "All projective sets have the Ramsey property" from just ZFC (like for the Baire property).

Hi David, This is my first foray onto MathOverflow as well, so this answer is an experiment to see if I can get things to work, rather than an attempt to convey a lot of serious information. As Andres said, the problem is still open as far as I know. I worked on this with Shelah a bit in the late 90s and we generated many things that led nowhere. Some related material: 1) Roslanowski and Shelah investigated "sweetness" and "sourness" (properties of ccc posets motivated by the constructions in Shelah's "Can you take Solovay's Inaccessible Away") in a series of quite technical papers early in the 2000s. This was partially motivated by the problem of getting all nicely definable sets to be Ramsey without using an inaccessible. 2) CH + "every set of real in L(R) is Ramsey with respect to every Ramsey ultrafilter" is equiconsistent with the existence of Mahlo cardinals. Mathias got the consistency result assuming a Mahlo, and the other direction is in a paper of mine from 1999 or so. The trick I used didn't seem to shed any light on whether or not the inaccessible is needed when we drop the reference to ultrafilters. Best, Todd 


Hi David. This is still open, and I don't know of any strategy that would result in a model with the property for all projective sets but not in a model with the property for all sets in $L({\mathbb R})$. Carlos Di Prisco has worked on this problem, you may want to contact him. Other than Carlos, the person to contact is Andrey Bovykin. He has been working on a project involving Harvey Friedman's "Boolean relation theory" whose goal is to establish the consistency strength of the Ramsey property. Unfortunately, I do not have any additional details on what his approach involves, and am curious as well. 

