Document the boostrapping strategy

[japaric]

Document the boostrapping strategy

[japaric]

The, perhaps not too explicit, plan is to be pure Rust from beginning to end (we are already doing this) and to incrementally provide an API that's 1 to 1 equal to std.

The holes in our API will be filled on an "as-needed" basis. Using the method described here, one can already start compiling crates against steed without changes, in the case of rlibs, or with minimal changes to main, in the case of executables. Obviously, most crates won't compile because steed API is not on parity with std but by trying we'll get compile errors that will inform us on what API need to be prioritized.

Is it a requirement that no external assembler

Not a requirement but I would prefer it because it helps with goal of "hassle free cross compilation". (I'm assuming one needs to install a different gas to compile e.g. ARM assembly into an ARM elf object). I'm not opposed to providing a Cargo feature to use gas instead of asm!.

[briansmith]

I'm not opposed to providing a Cargo feature to use gas instead of asm!.

I imagine it would become burdonsome to keep two copies of assembly language code in sync.