Well it certainly satisfies my architectural concern. However, I wonder if there's a useful precedent to set here w.r.t. the names of properties that are intended to be used this way e.g. prefix with a '.' or '__' -- a shorthand for the very low committment level.
tim * Chris Horne <Chris.Horne at sun.com> [2007-06-01 09:02]: > Tim Marsland wrote: > > Why do we need *both* the cb_ops flag and the property? > > > > If you're having the property, what are the semantics if > > the property is only defined on a particular instance, rather > > than the driver? Also, if the property is changed/removed by > > the driver, when does the framework notice? > > > > Suggest that rather than invent answers to these questions, we > > should just use the cb_ops flag - it seems to be enough to express > > this particular, more structural, behaviour. > > Would changing the commitment level on the property to "volatile", and > not delivering property man page changes address your concern? > > The motivation for the property was to provide relief in the field if > the new exclusion behavior causes problems. Getting a new driver with > a patched cb_ops flag to the customer in a timely fashion, especially > from a third party driver vendor, could prove difficult. Providing the > property implementation as volatile still gives us a mechanism is help > diagnose the problem and provide temporary relief until the cb_ops flag > driver patch is available. > > -Chris
