On Thu, Nov 27, 2014 at 5:12 PM, Robert Ramey
People using older compilers are mostly maintaining old code. I doubt they really need new libraries.
I hope that presumption doesn't drive any policy decisions regarding the future of Boost. Our product uses just short of 50 third-party library packages (one of which contains several Boost libraries). The ABI problem means that in practice, to upgrade to a newer compiler, we must rebuild all those packages with the new compiler. I need hardly mention that the idiosyncratic build scripts for each package need iterative tweaking to produce clean builds with the new build system. That's quite aside from the C++ source changes required by compiler evolution. Given that, I am very happy that this year my organization committed the resources to update to newer compilers on Windows and Mac. It's taken us three months. Although I personally consider our code base fairly large, I realize that a large organization would consider it laughably trivial. It's all too easy to understand that with a larger code base, or different management, such an effort might never have gotten a green light: we'd have remained stuck on pre-C++11 compilers. Yet our product is under active development. We continue to add new features and refactor existing code. We recently added a couple more third-party libraries. I am keeping a keen eye on the Boost library suite. I do not believe that Boost should require C++03 support for new libraries. As has been pointed out, the barrier for entry to Boost is already quite high. I'm just contradicting the assumption that those who must work with old compilers aren't doing any new development, or have no interest in new libraries.