I was hoping for more discussion from the mailing-list about this. I went in the opposite direction to always allow a 32bit model and to clamp down on use of strict 64bit models. (one 32 to many 64) I noticed others tried to use the "hardware matrix" in a top-down build direction, but that only speeds up the conveyance of such data. It is unclear, in multitasking, when to apply the "boundaries" before a player guts the existing data. In threads versus memory models, it matters, and C11 will get us much further than C++. If there is interest for more C11 details here than (in) a new paradigm for C++, please let me know.