Darren Reed writes:
> Please explain to me how having some value that is used in IP
> is worse off for being used via a variable (that can be tuned.)
> Exactly what is this "baggage"?

Having a tunable that's never actually used just wastes space and
promotes confusion.  Customers often *do* tweak random variables
without actually understanding what they do, and then get themselves
into all sorts of trouble as a result.  The output of the magic "?"
variable in ndd is just too much in their faces.

I'm not saying we should Windows-ize everything and make the system an
unmanageable black box.  That'd be dumb.  But we should also look at
what we're providing and see if it makes sense.  If it doesn't, then
fix it, where "fix" might mean making the tunable usable with a
different granularity (some are system-wide but should perhaps be
per-socket instead), or perhaps just nuking it if it's pointless.

That's the thinking underlying both the original dladm and Brussels
work, as well as this new work.

Since ndd cruft carries a lot of baggage with it (a noticeable bit of
code and often design consideration for new features as well, in
addition to a lack of any stable way to make changes that survive
reboot and lots of admin folklore), attacking these makes sense.  In
many cases, the "problems" that the variables were addressing either
have long since left the planet, have been overtaken by other
technology (such as ICMP rate limits), or just never existed at all
(many seem to have been added on spec).

Perhaps in *some* cases, it may make sense to "demote" some of the
variables from ndd to /etc/system rather than just converting straight
to a #define.  Yes, I realize /etc/system is system-wide and thus
might not be what you want with exclusive stack instance zones, and
that it lacks range controls.  So what?  If we're talking about a
variable that's "never" used in practice, and we're retaining it just
for future emergencies, then supporting exclusive stack instance zones
is just an unnecessary elaboration.  Plus, relegating it to
/etc/system automatically fixes the surviving-reboot problem, and the
in-the-admin's-face problem, so it's a positive step.

In other cases, just deleting would be better.  Like that VHF fine
tuning knob.  You ain't gonna need it.

-- 
James Carlson, Solaris Networking              <james.d.carlson at sun.com>
Sun Microsystems / 35 Network Drive        71.232W   Vox +1 781 442 2084
MS UBUR02-212 / Burlington MA 01803-2757   42.496N   Fax +1 781 442 1677

Reply via email to