Re: [fuse-devel] PATCH: Unmount on inactivity (2)

[Miklos S. <mi...@sz...>]

> > For sshfs, a better solution would be to disconnect from the server,
> > but not actually unmount.  And reconnect on new activity.  The
> > reconnect part is already working, and adding forced disconnection
> > does not seem too hard.
>=20
> Well, this doesn't fit my usage of network mounts. I use them to access
> data files, so basically I open what I need, work on it, save it, and
> close the application. I find it useful not to have to mount the networ=
k
> share manually (the automount feature), and not to have to remember
> unmounting it after I'm done (the unmount-on-idle feature).
>=20
> And I really want it to unmount when I'm done, as I am using a laptop,
> and there might not be anything to connect to anymore later (e.g. no
> network connection available).
>=20
> > This would have the advantage, that CWD within the mountpoint wouldn'=
t
> > defeat disconnection.
>=20
> The way I use it, a CWD within the mountpoint *is* activity, and I
> really *don't* want to have an unmount in this case. It seems to me the
> two features (disconnection and unmount-on-idle) are pretty orthogonal.

OK.

> > Right.  What I recommend, is that we wait a bit with adding this to
> > the library, to see if people find it useful, and possibly until a
> > stacking infrastructure is added to libfuse.
>=20
> I understand. Judging from the (lack of) feedback I have received to th=
e
> patch (except from Valient and Csaba), either you are right and only fe=
w
>  people find the feature useful, or there are no developers on the list
> that would patch their installation of fuse, and most people just use
> the releases. We'll see.
>=20
> What's that stacking infrastructure you mentioned?

There's a patch from V=C3=A1clav J=C5=AFza here:

  http://fuse.sourceforge.net/tmp/fsfipi.diff

but I think integrating it will have to wait until after 2.6.

Miklos