i just realized i disagree with the below but here's what i had typed i guess i mostly am attracted to the idea of a toolchain that "just works" everywhere because that in itself is cool. it also means we come in to the next phase, where we differentiate native build products through e.g. variants, with a deep understanding/codification of the various types of binary compatibility issues, which we wouldn't get if we just tossed an image out there. i also feel like if we can't solve this on the build side we're going to have a heck of a time on the consume side, where the problems are basically the same, unless we also just produce docker images or equivalent from our native code instead of shared libraries. i think seeing how effective pex is and how facebook has already tried out a major tool with FUSE (xar) the idea of providing some sort of stable environment sounds great -- as long as we can still produce shared libraries / etc so we can make e.g. conan packages. i view the current work as being solvable and having that done and making it work reliably gives us a basis for integrating new languages and workflows into native code. it feels to me like not solving this problem now means we'll have to solve it slowly, later, and the upfront work here feels like the right tradeoff (to me, right now).