Cedric Blancher writes: > On 21/06/2009, James Carlson <carlsonj at workingcode.com> wrote: > > If anything, moving bits out of *both* SUNWesu and SUNWcsu and into a > > clearer-headed "SUNWksh93" (or some such) would make more sense. > > Which sense does it make to have a separate package which makes a > machine unbootable if you pkgrm it?
First of all, I'd dispute the "unbootable" assertion. It might well be perfectly reasonable for someone creating an appliance out of OpenSolaris software to create a system that doesn't have ksh93. Perhaps such a system only has bash! But more importantly, I think what you're suggesting is that we should have one giant package with everything needed to boot. That'd be silly. Why should I have to upgrade the entire system when some little patch is released for ksh93? Packaging isn't only for selective installs (though that's an important function); it's also a way to set boundaries on software that's maintained as a single unit. In the minimum metacluster (SUNWCmreq) for Solaris, we currently have 18 clusters of packages, plus around 73 other packages. We already have lots of software that you "can't remove" and that's nonetheless delivered by separate packages. The simple metric you're suggesting (needed to boot -> single package) is not and has never been the way Solaris software has been designed. -- James Carlson 42.703N 71.076W <carlsonj at workingcode.com>
