Re: ln -P/-L and linkat

[Jim Meyering]

Eric Blake wrote:
> Now that POSIX 2008 requires ln to support user choice of whether to
> create hard links to symlinks, we can probably get rid of
> ENABLE_HARD_LINK_TO_SYMLINK_WARNING in the ln.c source code.

Yes.  Good point.

> I noticed
> this while considering how to implement linkat in gnulib.  I guess
> platforms that follow link() by default will just have to fail with ENOSYS
> since they don't support -P (and keep their default of -L), while systems
> like GNU/Linux that have always behaved like -P are now justified.

Agreed.

> Do we
> want to make ln's behavior conditional on _POSIX2_VERSION (where
> conformance to 2001 tries to imply -L, even on Linux where it is more
> natural to imply -P), or is it not worth worrying about?

Let's not worry about this.
That will be clearer, and besides,
I doubt anyone would want to flip that switch.