Re: [Utopia] gnome-vfs patch, take one

[David Zeuthen <david fubar dk>]

On Thu, 2004-05-13 at 20:41 +0200, Sjoerd Simons wrote:
> I'm always wondering how gvm is going to work out, when your various people
> are simultaneously logged into one machine and all running gvm. Using
> X-terminals for example. Multiple gvm's racing for a hotplugged volume, volumes
> being unmounted when one of the users logs out.. But i don't know if that
> really falls into gvm's scope... 
> 

The easy answer is that only the g-v-m instance(s) for the user at the
console will have sufficient privileges to mount - this still means a
race if you're logged in at the console and at a remote terminal. I'm
not sure of the implications of the races from this, but it would
probably be sane to only run g-v-m for sessions at the console.

But your question is perfectly valid - now add some USB port to your
remote displays and users will want to attach god-knows-what devices to
them and use them on the computing server running their session. Which
is a perfectly valid desire. IMO, it's a complete research project on
it's own to do such things. However.. I do think that even this may be
within the scope of Project Utopia, at some point in time.. Oh well, I'm
rambling again; sorry :-)

> > > Also i don't think it's nice to ``force'' people to use the command line,
> > > just because they don't use gvm or have disabled automount. But i'm not the
> > > one writing the code :)
> > > 
> > 
> > Anyway, I can see problems having icons for unmounted media. One of them
> > is that we want to teach users to click unmount media before yanking it
> > out of the computer 
> 
> I tried teaching my parents to use click umount before pulling out their
> camera, but it just doesn't work. They forget about it and just yank it out and
> untill now they've never had problems with it, so there is no reason to
> remember it for them..
> 

So do I (when there is no IO) and HAL will do a umount -l /dev/whatever
and things work just fine. Kernel people, though, will tell you this is
unsafe and in fact I crashed 2.6.5 last week while accidentally doing it
while there were IO.

To me, it's still an open question if g-v-m should put up a huge banner
notifying the user that they did the wrong thing.

> > - it would just feel odd if the icon didn't disappear, wouldn't it?
> The icon should disappear from the desktop, but it's nice if there is an icon
> in the ``computer '' when there is a volume available but not mounted. Which is
> also the current behavour.. 
> 
> On the other hand, i don't want icons there on things that i can't mount anyway
> (stuff like an apple bootblock and partition map for example). But that's a 
> more general problem, because there is currently no way to know before hand
> what can be mounted or not afaik. For example the fstab-update callout, also 
> shouldn't really put things like that in my fstab.
>

Actually HAL now detects the fstype (thanks to Kay Sievers) on the block
device, so in fact you can enhance the fstab-update script accordingly.

Btw, this is exactly why it's implemented as a callout, because it
varies by OS/distro what filesystems are supported in the kernel - e.g.
european distros may ship with ntfs support while us-based ones wont.
 
> > It's not that I'm totally against changing this behavior, I just see
> > little gain and potential usability problems. And in certain cases, my
> > experience tells me that the person writing the code should be the last
> > one to decide policy and defaults :-)
> 
> When someone indeed is using gvm with the proposed automounting behaviour, 
> there isn't much gain. I just wouldn't want to force people into using gvm with
> automounting enabled.
> 

Maybe the gnome-vfs stuff should check for the gvm gconf value and
implement the behavior you are suggesting.

Cheers,
David