Re: [PATCH] close windows on unmount

[Ryan McDougall <ryan mcdougall telusplanet net>]

On Fri, 2004-03-12 at 09:50 +0100, Alexander Larsson wrote:
> On Fri, 2004-03-12 at 07:36, Ryan McDougall wrote:
> 
> > Special casing this way doesn't look quite right to me. I think the
> > problem is not at the nautilus level, but at the file system abstraction
> > level.
> > 
> > Perhaps we need a new signal: "unavailable". When the FS unmounts, it
> > implies that resource is gone from the system, and unlikely to return in
> > its present form any time soon; so nautilus window closes. When the FS
> > is unavailable it means that the resource is not reachable, but is still
> > there when you ask again, provided that "connection" (in what ever form
> > it may be, whether automounted or network connected) is re-established;
> > so nautilus keeps the window open, and tries to "reconnect" when an icon
> > is clicked. If it is unable to reconnect, then it errors out and closes.
> > 
> > That way the two signals capture the semantic difference between "I'm
> > gone" and "I'm temporarily unavailable".
> > 
> > I think what you propose is a bit hackish. Comments?
> 
> And practically, how would you implement this on a unix machine?
> 

Perhaps this is the point where you realize I'm in over my head,.
However, I do *not* suggest that we somehow change UNIX, rather I think
there is some place above the kernel/FS layer where this can be
implemented, specifically gnome-vfs. 

I'm sorry I can't be more help, since I am unfamiliar with how nautilus
currently handles OS level things such as file descriptors.

Cheers,
Ryan