As someone working in high-performance systems, I am all too keen on the ability to avoid heap allocations wherever possible, whether for cache-locality or lower-overhead. The news that custom allocators where being rolled-out, in nightly, was therefore of great interest. See Box and Vec. Unfortunately, the design of custom allocators prevents inline storage -- that is, a Vec<T, ???> cannot be sel
![Is custom allocators the right abstraction?](https://cdn-ak-scissors.b.st-hatena.com/image/square/71479b5ecc3036c547c378fdc8815840708b4229/height=288;version=1;width=512/https%3A%2F%2Fglobal.discourse-cdn.com%2Fbusiness6%2Fuploads%2Frustlang%2Foriginal%2F2X%2Fb%2Fb71959a2097ce0b5c3a193ce0a6466a42d4c4952.png)