In your particular case of {x,y}, I'm also ambivalent. Influential people like Nevin Liber (P1163) have expressed strong aversion to HORSE, so, to a first approximation, HORSE will never happen. Also, people today are physically prevented from doing CART. Your proposal proposes to permit people to do CART, knowing that HORSE will never happen.

So there are (at least) two possible positions: Nicol's "I don't want CART without HORSE" and my "well, we obviously can't get HORSE before CART, so we might as well get CART and see if that helps the HORSE to follow."

I am not personally making any prediction whatsoever on whether we will ever get HORSE or not; I do not have any secret agenda! :) If this kind of proposals are useful to spur old discussions, by all means, let's have them.  (But in another thread :-P. By the way, P1163 got rejected).

My *strong* position is: there are plenty of legitimate, non-niche use cases for initializer lists that don't have the HORSE problems. These initializer lists are already usable with certain stdlib functions, but they're not usable with other functions (sometimes strongly related, e.g. push_back/erase) because of an oversight. My aim is to simply close that oversight. I definitely do not believe that improving this CART is going to increase the adoption of "wrong" usages so much that this will effectively block any future tentative of fixing the HORSE.

Thanks for your feedback, stay safe,

-- 
Giuseppe D'Angelo | giuseppe.dangelo@kdab.com | Senior Software Engineer
KDAB (France) S.A.S., a KDAB Group company
Tel. France +33 (0)4 90 84 08 53, http://www.kdab.com
KDAB - The Qt, C++ and OpenGL Experts