With regards to endianness of the web: this goes back to the original Khronos spec for typed arrays where the only view was machine endianness. At least now the

With regards to endianness of the web: this goes back to the original Khronos spec for typed arrays where the only view was machine endianness. At least now there are both machine, big and little endian views… Exposing machine endianness was largely criticised by Mozilla and Opera people (AFAIK, only the latter were still really shipping on big-endian platforms). It was largely motivated by a desire to ensure WebGL could have good performance when copying data on all machines. Of course, plenty of us saw how that would end: everyone testing only on their hardware (which would inevitably be little-endian), and nothing working on big-endian hardware. Sadly, the majority of the WG didn't really care, given the only big-endian stuff they cared about was things like TVs where they only cared about (web-platform based) widgets, and they only cared about the WebGL case, and didn't believe anyone would use them for anything else…