Peter Memishian wrote:
> > I think the point that I hear Peter making is that one can imagine
> > changing the "default" value for a property.
>
> Yes.
>
> > Thinking about this somewhat further, I have my doubts about the utility
> > of a writable "default". As I see it, Brussels properties are now
> > persistent. Why would an administrator change a "default" independently
> > of changing the lasting persistent ("non-default") value? Its just not
> > clear in the context of Brussels what the point of maintaining a second
> > persistent value is.
>
> The idea would be to provide a general default value that would be more
> applicable for their configuration, but then still provide an ability to
> override it persistently on a per-link basis if needed. As we bring more
> properties into the mix, I could see some meaningful combinations -- e.g.,
> default Wake-on-LAN settings that one might want to configure across all
> applicable links. Existing properties such as the WiFi radio or powermode
> settings may already be things the administrator would like to set a
> different persistent default value for.
>
Okay, except that Brussels currently only has per-link tunables.
Meaning there is no master repository of "defaults" for all links, or
all links of a certain class. I believe the desire for system-wide
defaults may have merit (not entirely convinced, and I see problems with
it), but it definitely falls outside of the scope of what we're talking
about here -- which are defaults for a *given instance* (link), and
which come from specific hardware in question.
I *think* we're approaching convergence here. I'd be happy to see
DLIOCSETPROP with DLD_DEFAULT removed from the spec, not worrying about
atomicity, and keep it simple for now.
We can explore other ideas, such as system-wide defaults, later.
(Btw, the idea of system wide defaults is yet another reason why ndd is
badly chosen for data links -- it simply wasn't designed with their
needs in mind. For IP, you normally only have a single global IP stack
-- ignoring IP instances for the moment -- so the idea of a system wide
-- or stack-wide at least -- default makes sense, with possibly
different per-interface overrides. But for link-layer tunables no such
"master" set of tunables exists, anywhere, in Solaris today.)
-- Garrett
