Re: sketching a mountpoint naming scheme (was: [Utopia] Re: NTFS permissions are wrong using fstab-sync?)

[David Zeuthen <david fubar dk>]

On Thu, 2004-08-19 at 14:41 +0200, C. Gatzemeier wrote:
> > However, I don't think both features matters much eventually; the UI
> > shouldn't be exposing the mount point name and it should use the UDI as
> > the stable reference to the media. But that's something for the
> > future :-).
> 
> Are you also considering CLI or older ncurses apps as UI? I'd appreaceate nice 
> mountpoints that work system wide, depending less on desktop-level-only 
> abstraction.

That would be nice as well, sure.

> Here is a preliminary idea for a mountpoint scheme, you'd know how fstab-sync 
> can relate to this:
> 
> (1)
> Devices considered or subsequently defined as static, say those available 
> during installation (or boot?).
> 
> -- standard removable media devices -> /media/floppy (cd, dvdrw etc.)
>    (type-names)

fstab-sync does this today for devices that are likely to use media that
isn't partitioned, e.g. floppy and optical drives 

(look at the hal property storage.no_partitions_hint; it's a hint only
because you may actually put non-partition memory cards in a CF card
reader).

> -- otherwise not mounted partitions -> /media/<volume name>
>
> (2)
> new devices (especially hotpluged):
> 

We do export a property on the drive called storage.is_hotpluggable that
can be used for this.

> -- removable media devices -> /media/<id string>-<type>
> 
> -- partitioned devices -> /media/<id string>/<volume name>
>
> 
> 
> <id string> is the first available of model-name, vendor-name, "humanized"
>                   bus id ("ide-drive-#" (i.e. a=1,c=3),  "usb-id-#" etc.)
> 
> <volume name> is the first available of volume-name, "humanized" partition id
>                   ("partition-#")
> 
> <type> are things like "floppy", "cd", "dvdrw" "storage"(usb) etc.
> 

We export all of this information, and more, as properties in hal.

> ----
> 
> The following from the previously posted current example seemed particular 
> disturbing to me ;-)
> 
> /dev/hda8               /media/idedisk  
> /dev/hda6               /media/idedisk1 
> /dev/hda1               /media/idedisk2
> 
> What is the second enumeration based on? What's its purpose?
> 

It's just how it is right coded right now and the enumeration is totally
unstable (e.g. the mount point name is not stable) and not really that
well thought out.

(If you want totally stable mount point names but don't care about what
they look like you can actually use the hal udi for starters (ok we may
need some more work on it but it will be as stable as possible).

Personally, I'm a bit wary about encoding too much stuff into the mount
point name; some of my personal preferences includes no spaces, short
names and few underscores in them. But that is just my opinion.

I'm also not super happy about creating a configuration file for fstab-
sync, it's just another attack vector, however it seems like a
reasonable compromise given the fact that people have so different
opinions. Care to write a patch?
 
Cheers,
David