Re: File type nomenclature in 170 Lassi Kortela 31 Jul 2020 15:45 UTC

Thank you!

> I'm fairly indifferent about "fifo" vs. "named pipe", but note that
> Windows Named Pipes have very different semantics: you can only create
> them in the magic named-pipe file system, they are visible across the
> network, and they go away when closed.

I thought "named pipe" is a generic OS concept, not just Unix and
Windows. But maybe I'm mistaken. Does e.g. VMS or Amiga have them?

> I very strongly prefer the predicates to returning a magic symbol,

I don't have a strong opinion one way or the other, so predicates are
okay. Is the rationale for predicates that their names can be checked by
the compiler/evaluator, or that predicates can handle types that are not
disjoint from other types (i.e. a type is a superset of other types)?

Magic symbols do have this benefit:

(case (file-info:type foo)
   ((file) ...)
   ((directory) ...)
   (else ...))

> I agree with changing "if file is special" to "if file is a device".

Do you agree with merging block and character devices into "device"? The
distinction is esoteric, and the few people who truly need to care about
it can look at the raw bits in `file-info:mode`. Windows also has a
"device" file type but not character or block.