In message <[EMAIL PROTECTED]>, Shaya Potter writes:
> Erez Zadok wrote:
> > In message <[EMAIL PROTECTED]>, Shaya Potter writes:
> >> Erez Zadok wrote:
> > 
> >> why not make it a device node with commands that can be issues to it. 
> >> like an ioctl, just different (better?)
> > 
> > Where will the device live?
> > 
> > - if in devfs, then we depend on devfs to exist
> 
> devfs doesn't exist anymore.
> 
> > - if in /dev, then we need the users to MAKEDEV
> 
> so bad?

Yes.  Sysadmins don't like to have to know to mknod extra nodes (and they
might forget to delete the node if they want to remove unionfs from their
system).  Then you have to worry about major/minor conflicts.  And, forget
about trying to get any major OS vendors to update their /dev/MAKEDEV to
include unionfs entries.

> > - if in the mounted union, then how different is it from a "magic file"?
> 
> it's not, I was thinking more along the lines of not having to parse the 
> whole thing in userspace as you say.

I assume we'll have a libunionfs which'll include useful wrapper functions
to read in a config file, add/change/swap/del entries, and write it out.

Erez.
_______________________________________________
unionfs mailing list
[email protected]
http://www.fsl.cs.sunysb.edu/mailman/listinfo/unionfs

Reply via email to