Quoth David Powell on Thu, Apr 27, 2006 at 02:42:28PM -0700: > On Thu, Apr 27, 2006 at 02:04:19PM -0700, David Bustos wrote: > > Quoth Stephen Hahn on Thu, Apr 27, 2006 at 12:58:49PM -0700: > > > * David Bustos <David.Bustos at sun.com> [2006-04-27 11:45]: > > > > Hmm, I wonder if we should also allow svcprop with no operands, which > > > > would automatically use scf_myname() to determine the current service. > > > > > > Requires all processes to provide sanitized environments to their > > > descendants or scf_myname() to be a bit more contract aware... but > > > interesting. > > > > I don't think that's a requirement; we can just say "Only guaranteed > > inside a method process. Will work in child processes as long as the > > environment is not abused." > > If we're talking about the difference between working and getting a > usage message, I think we'll need a much stronger definition of the > context in which the command works. Even then, I'm not sure I like > the idea of a normal-looking command whose behavior could change > based on "invisible" inputs. I'd be much more at ease if this > behavior was turned on with an option.
I think it's a stretch to call svcprop with no operands "normal-looking". I think it would be safe to nail down the context as "where $SMF_FMRI is defined". > Another side effect of this would be that methods couldn't be run > directly, i.e. outside of SMF. This normally isn't -- and shouldn't > be -- a problem, but it does make debugging harder. That's already true of methods which use $SMF_FMRI. David
