> However, better break things early. When can we expect the transition of io > library? The sooner the better.
This doesn't buy us much by itself, so we had not planned to make it an immediate priority. I expect to have a quasi-usable debian-hurd distribution before we get around to it. (Switching to libio requires a soname change, and when it will really buy us something is when we implement pthreads, which will also require a soname change. Pthreads is not likely to happen real soon, so if libio happened first there would be two soname changes by the end of it.) On the other hand, it's probably not all that much work if someone wants to volunteer to do it. The first step is to port libio to the hurd; this can probably be done trivially, i.e. just compile libio as it is and it will use the POSIX.1 open/read/write/lseek functions. Currently, the hurd's sysd-stdio.c is set up to do something more efficient using the hurd_fd data structures directly; that would be good to do in libio, but it's not essential. That should be enough get purely Unixy stuff going with a libio-based libc. Then convert the stdio-magic stuff in libc: mach/devstream.c, hurd/vpprintf.c, hurd/fopenport.c. Finally, the exec server uses stdio magic that will need to be converted to equivalent libio magic. This is probably the hairiest thing that needs to be done (in exec/exec.c). For someone familiar with libio (I'm not very), this can all probably be done pretty quickly.
