From: John Cowan <xxxxxx@ccil.org>
Date: Monday, August 26, 2019 1:35 PM

On Mon, Aug 26, 2019 at 2:21 PM <xxxxxx@ancell-ent.com> wrote:

We currently do allow ports and necessarily use fstat. We could drop ports as a possible argument to file-info, that would make the SRFI simpler to implement,

Not much, since it's just a call to port-fdes.

I recently noticed this can be made automagic in Chibi Scheme's FFI. But it's still an extra C wrapper, which we've considered to be significant enough to drop some other procedures from the SRFI. But as you note, worth it:

Which poses a problem if we make it required across the board

Again, not much. A port can't be open on a symlink, so you'd pass #f. If you pass #t it makes no difference, but then it makes no difference 99% of the time. Having fstat capability, though, helps eliminate race conditions from stat-before-open.

(and we should decide RSN if it's going to be chase? or follow?).

Follow. "Chase" is just Olin's cuteness.

Will do, and will add follow? as a required argument, to be ignored if file-info is handed a port.

- Harold