Messages in this thread | | | Subject | Re: Devfs, was Re: Migrating to larger numbers | Date | Wed, 9 Jun 1999 01:01:22 +0100 (BST) | From | Alan Cox <> |
| |
> Ultimately I'd like to see proper volume-based mounting. Not UUIDs (as > Linus said to me privately, 128 bit numbers are ugly), but proper > names. Like "root", "usr", "local", "home", "data", "tmp", "swap" and > so on.
That doesn't work. You need the uuid as well, but the uuid stuff can be buried out of user sight in general. The reason you need uuids too is simple.
You take some machines down for service. You plug the disks back in. You boot the machine you are rebuilding. You reformat /database. Then you realise you had the fibrechannel cables crossed and you reformatted the other systems /database.
That is why UUID's are essential and the more common less destructive case of wanting to stick another hosts /home on your chain for some work, and the fibrechannel fabric case where there are multiple root home and swap partitions visible on the same fabric.
The UUID is there to be sure you get the right volume. Its not there to be inflicted on the user directly. I should be able to mount "home" and be told "which home" or "the only home I can find isnt the one you mounted last time"
If you want bonus points then you can RPC for a remote host with the volume on the local network for NFS ;)
> I'm still debating with myself whether volume names should be > registered with devfs. In other words, the code that determines volume > names registers them under /dev/volumes/
Interesting idea.
> Are your concerns related to the devfs concept or to the particular > naming scheme?
The concept.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |