On Tue, Feb 25, 2003 at 02:42:02PM -0800, Chris Wedgwood wrote:
> On Wed, Feb 26, 2003 at 09:06:28AM +1100, Nathan Scott wrote:
> > The only thing that could go "wrong" is in xfsdump/restore at the
> > moment, and there are acceptable workarounds (IMO).
> xfsdump/restore are pretty important IMO
> these are not applications i'd like to see trivially broken...
Absolutely. They aren't "trivially broken", they just need to
be told if you want to do things the old way (any attempt by
restore to set an attribute on a non-existant namespace will
fail & produce error messages of course).
> is it not possible to have userspace deal with both cases?
Userspace does handle both cases, it just can't do both at once,
and it can't intuit which to use either (as you suggest below).
> it on'y
> needs to determine which namespace prefix is in use once and cache
[there may also be no namespace prefix in use (yet) ... if no
ROOT names are set at that point in time]
> this result (somewhat like what glibc does in places)
Yes, that would be good - I can't see a good reliable way to tell
which namespaces a particular kernel supports though.