On (10/05/07 06:52), James Carlson wrote: > > Sowmini.Varadhan at Sun.COM writes: > > Did you want to see this in the 20q.txt (some of it is there in > > answer to the question "what are the clients over which a change > > should be managed", but I could provide more details if you feel > > it is needed (without confusing the deliverables associated with > > 2007/429 itself). > > Either the 20q or the design would be fine. It didn't seem clear > enough to me that we're trying to get rid of ndd's use for this > purpose, or how we're telling users to make the switch. > > You could do something like this in the interface table: > > ndd on drivers Obsolete replace with dladm > > ... and then include an update to the ndd man page and the release > notes that indicates what users should do.
Ok, I will include the above two (ndd.1m related man pages have been updated - these already come with few guarantees and many disclaimers - the Release Notes update to appear soon). > Bonus points for making > ndd issue warnings when used in the now-obsolete manner. We were going to do this in each driver as it was converted (e.g., the ndd handler in bge will print out a message such as the one used for the ndd usage to set ip_forwarding as part of PSARC/2003/166). > > I think the questionable part here is actually the timing of this > message and exactly what the message will say. Perhaps it's too early > to be directing people to use dladm instead of ndd when this project > integrates ... ? We can minimally emit warnings for all the MII properties of each converted driver after 2007/429. The other properties (e.g., drain_max) are typically hacky outliers, that can be converted to private properties (as shown in Appendix B of the design doc), and when that is done, we can still emit a message for these in the ndd path. > The design document seems to focus on the mechanics of making the > compatibility bits work. That makes sense, but it doesn't explain > what we want users to do. Ok, so that makes me feel that the 20q look like the more appropriate place to address the transition strategy. I just updated the 20q.txt with this information (see response to Question 2). --Sowmini
