bug#26126: 26.0.50; file-notify-rm-watch removes arbitrary watches

[Eli Zaretskii <eliz@gnu.org>]

> From: Michael Albinus <michael.albinus@gmx.de>
> Cc: politza@hochschule-trier.de,  26126@debbugs.gnu.org
> Date: Wed, 22 Mar 2017 14:23:47 +0100
> 
> > But Andreas asks about calling remote handlers, about which we by
> > definition know much less.
> 
> Nope. We know exactly which remote handlers are called, and how they
> behave.

We know that about our handlers, yes.  But that doesn't have to be the
end of the story.  Emacs is extensible.

> Do you expect other implementations of remote handlers?

Yes, why not?  It's much easier to do that in Lisp than in C, where
the local handlers should be implemented.

> > In that context, it might indeed make sense to pass the file, not its
> > parent directory, because the handler can easily reconstruct the
> > parent directory if that's what it needs.  By contrast, there's no way
> > for the handler to intuit the file which was stripped.
> >
> > WDYT?
> 
> I still don't understand what's the difference between local and remote
> events in your eyes.

See above.  Admittedly, this is a minor point, so not worth arguing if
you disagree with my POV.

> I've tried to implement remote handlers to behave exactly like the
> local ones. That's the Tramp philosophy.

Right, but in this case there are 2 flavors of local handlers, and the
question is on which of them to model the remote ones.