Erik O'Shaughnessy wrote: >> On Aug 6, 2008, at 4:00 PM, Nicolas Williams wrote: >> Erik: do you need a procedure for getting kstat version information
> Yes, this is a deficiency in my design and I will address it. I I'm not sure where the value is in this... There is "what version of kstat do we have?" and "what kstat-able data resides in the system?" The first is somewhat meaningless, and there is no simple answer for the latter - even an enumeration of the kstat-able value names tells you nothing about the implementation-details of what the data referenced by those names actually means. Yes, there probably semi- to completely-manual heuristics that the client code author could use to determine whether the data is usable, but no matter what they did, they would be building a house of cards. When the Explorer cases went thru, we debated this topic to death - and surmised that the only thing this kind of data could be predictably used for was collection for human perusal - any automated use would require interaction with the team(s) that created the kstats themselves to upgrade their stability levels. -John
