[Apologies for the cross-posting.]

I've kept out of this discussion up until now, but would like to break
my vow of silence by simply calling for a timeout period.

Also, many thanks to Jim Wise for an excellent article which prompted
this discussion.

The reasons for the desired timeout are many:

1.  We've just finished a period of consolidation in pkgsrc, and now
everything seems to be happening at once.  I really would like to get
on top of things before we start modifying pkg_* to do anything (Those
of you who know me will accept that I'm a great believer in the "one
step at a time" approach, and I do feel that we're reaching a bit
beyond ourselves here).

2.  I want to bring both my i386 boxes up to -current, and what with
the hiatus of everything in the last two to three weeks, I've not been
able to do that.

3.  I'm not convinced that we're all talking on the same wavelength
about this.

4.  I need to get the "Changes in pkgsrc in September" mail for
netbsd-announce done.

5.  I am in the process of moving jobs.

6.  There are not enough hours in the day. Or bars of chocolate.

7.  The pkg_* tools are in transition even as we speak, since Hubert
has some mods that are about to be committed, enabling the dewey
version relationships and alternate package dependencies throughout
pkgsrc, and some other mods, which will make the pkgsrc DEPENDS and
CONFLICTS definitions much easier.  In addition to that, and once
Hubert has committed those, I have some far-reaching mods I would like
to make to (a) bring some order to the pkg_* tools, and (b) move the
package databases from /var/db/pkg to be in ${PREFIX}/etc/pkg,
removing the PLIST @cwd directive on the way.  This will allow us to
eliminate the staging area from the pkg_add tools, which is a source
of great waste of cycles, brainpower, disk space, time and money. 
However, this may prove to be a step too far, and so I really want to
get a good idea of cost versus benefit, and move on from there.

With all the above happening, I feel that deciding how any system
package would look is a bit premature.  However, I do realise the need
for speed in this matter. Just not the need for haste.