Bart Smaalders wrote: > Dave Miner wrote: >> James Carlson wrote: >>> Dave Miner writes: >>>>> The other poster who mentioned patches on unbundled software was >>>>> on-target, though. Since that *is* possible, and since those sorts of >>>>> patches may not be going away soon, I think we have an obligation to >>>>> get pdo.c out there. >>>>> >>>> I think we have a Solaris product requirement to provide a utility which >>>> can apply those patches, and an OpenSolaris requirement to make the >>>> source for that utility available. Doesn't have to be pdo, could be >>>> something else. Which one we choose seems fairly bound up in the >>>> compatibility story, which clearly isn't baked yet. >>> If it isn't pdo.c, I'd wager that it'll take longer to release, as you >>> have to develop it first. >>> >> Might be. We might also just declare pdo the evolutionary dead-end that >> it looks like to me, and go back to the pure ksh code as a working base. >> That probably wouldn't take any more time to open-source, and I >> suspect could advance more quickly. >> > > You're going to teach the ksh scripts about zones? >
As I recall, pdo's knowledge of zones basically consists of computing the global dependency tree, and then turning around and calling the ksh script to effect the actual changes, which as we all know are primarily implemented in the sparse package scripts. From my point of view, it's based on a rather static model of zones attached to a system that doesn't match up too well with a view of zones as dynamic, mobile application containers, and it didn't even bother doing that very well, since the exit codes are a complete fiasco. Anyway, I don't regard the computations it does as being particularly exotic. Just my opinion, though. Dave
