On (01/24/09 14:30), Peter Memishian wrote: > > > > - I'm skeptical of the need to provide non-persistent object > > > manipulation. My assertion is that much of the system administration > > > model these days is implicitly persistent, and the overall system > > > usability would benefit from this being the default (if not only) mode > > > used by networking as well. > > > > I personally agree with you, but the design also does not prohibit > > the addtion of non-persistent versions down the road, should we > > find an actual need for these. > > I'm confused; how could libipadm be used by ifconfig without > non-persistent support? The ipadm utility should be very simple and > should care very little about persistence -- all that logic should be > in libipadm. I don't think the discussion is about where the persistence logic should be, but whether ipadm itself should support both temporary and persistent configuration. Mixing and matching flags can lead to the need for dependancy resolution where we dont allow permanent operations on temporary objects, so I had proposed in the document that until we see the need for it, all ipadm operations would be persistent.
As you point out, the converse situation holds for ifconfig- all ifconfig operations today are non-persistent, so the library really needs to be able to do both versions (in order to support both ipadm and ifconfig needs), though ifconfig is not going to start support the "persistent/non-persistent" flag. --Sowmini
