I took Thomas' email to mean the decision was Fernando's, and it would be him (Fernando) who would commit.
Indeed. I will commit sometime within the next hour.
I don't have strong opinions on whether we revert or apply my patch. Reverting will definitely break precompiled headers involving <boost/none.hpp> on VC7.1 and Borland; my patch contains Steven Watanabe's attempt to solve this (which I haven't been able to verify as I don't have access to either affected platform, though it compiles and passes its tests on other systems), and includes a test case for the original regression, but it could introduce a regression on some other platforms.
If Fernando thinks it's safest to simply revert, that's absolutely fine by me.
I don't have access to those compiler either, so there is some uncertainty in the patch, even if not w.r.t to regressions. So given that I can't test if it really works (in terms of making PCHs happy), I rather just revert. But of course I will apply it, along with the test case, in HEAD, for 1.35. Best Fernando