Subject: Re: major hier(7) overhauls?
To: Scott Reynolds <scott.reynolds@Plexus.COM>
From: Greg A. Woods <>
List: tech-userlevel
Date: 01/27/1999 14:03:17
[ On Tue, January 26, 1999 at 01:48:23 (-0600), Scott Reynolds wrote: ]
> Subject: Re: major hier(7) overhauls?
> What does `all in one file system' have to do with hier(7)?  It really,
> really sounds like your beef is with sysinst, not the fact that hier(7)
> defines the /usr file system.  If that is in fact the case, I have no
> problem with you wanting to have / and /usr on the same file system,
> provided that you don't force it on everyone.

Nothing really.  Did you miss the first part of this thread?

The issue was originally only to do with the fact that it's difficult to
make a package system hierarchy (eg. /pkg) orthogonal to the system
without having /pkg/usr, etc. or without flattening out the hierarchy
within /pkg and dealing with some rather inelegant hacks to make it
possible to have "/pkg -> /".  I pointed out that several systems,
including Plan 9 and the GNU Hurd seem to have successfuly flattened out
the system hierarchy too and eliminated the last vestiges of a rather
ancient hack called "/usr".

In the blue-sky world I think it would be good for all systems that are
evolving beyond what Unix originally was to eliminate /usr too.  Given
the close ties some (NetBSD) folks have with tradition though I don't
expect to ever see this happen to NetBSD, in particular.  Some people
seem to attach (too much) emotion to things that have merely technical
reasons for being (once upon a time they did anyway).

The other bits about whether or not this is possible on small machines,
and discussions over the releative merits and demerits of putting all
the OS in one filesystem are all side issues of separate importance.

I personally don't have any beef with having a separate /usr on NetBSD,
though I don't see that it gains most people anything at all, and in
some cases there's a noticable loss for having both filesystems on the
same spindle.  These kinds of issues really should be studied more
closely and documented better.  In an ideal world sysinst could ferret
key bits of information from the user which might help it make good
suggestions about how to gain the most resources from a given system for
a given purpose.

							Greg A. Woods

+1 416 218-0098      VE3TCP      <>      <robohack!woods>
Planix, Inc. <>; Secrets of the Weird <>