On Wed, Feb 27, 2008 at 11:31:34PM -0700, Sanjay Nadkarni wrote: > I am not sure I quite understand the "check for feature" concept. How is an > average user supposed to know that ? Take the example of Xvm. An admin is > told to install a version of Solaris that implements Xen 3.1. Using the > current approach, a sysadmin can perhaps map this to "Nevada bld xxx" or > in future Solaris 11 U1. How would this map in the "check for feature".
Depends on the feature. If it's a bugfix that's internal to some binary, then you're probably out of luck at install time, and have to resort to a feature -> version mapping (but that should be a version of the package containing the feature you care about, not the entire OS). But if the feature can be tested for by looking for a particular package, or a particular file, then that's easy enough, or should be. At runtime, there's no excuse for checking the OS "version", regardless of how you discover it. It's just too broad a brush for anything you might care about. Danek
