Ian Collins wrote:
> UNIX admin wrote:
>> As to subject says, I believe that the "small /, the rest for
>> /export/home" is inadequate as the default. What good is a default
>> if it ends up being changed as a rule, and not an exception?
>> 
>> What I'd like to see is the default to change to the "whole disk /"
>> approach, like IRIX and HP-UX have had for the past 17 years (and
>> proved that it works really well).
>> 
>> 
> That's pretty useless if you want swap or intend to use live upgrade.
> 
> 
> See the earlier caiman-discuss thread "default VTOC layout in dwarf 
> caiman" from late May.
> 

The Slim Install project will be making a proposal at some point in the 
next month or so about the disk layout with ZFS (we're not really 
interested in fiddling further with UFS layouts at this point).  But the 
general direction is towards more pooling and less slicing & dicing.

The one caveat, though, is that I think there's still some merit to 
having separate pools for the "system software" vs. "user data".  Even 
with ZFS, your fault boundary is at the pool level, so containing a 
system software failure separate from user data feels attractive to me. 
  There are also some limitations (at least for a while, see the ZFS 
boot case[1]) in the pool structures that ZFS booting will support, 
which also seem unattractive as constraints on the user data pool.

Dave

[1] http://www.opensolaris.org/os/community/arc/caselog/2006/370/


Reply via email to