On Wed, May 20, 2009 at 09:16:34PM -0400, Ben Walton wrote: > I suspect that this > evolution will need to happen anyway in the face of IPS, which won't > limit itself to the current rigid > dependency system.
funnily enough, there was some semi-related discussion on the ips dev list. And they're running into problems, because they decided to be "more flexible". So I dont think that just copying what IPS does, is a good general strategic move for us. Sometimes, offering more choices, actualy ends up being the wrong thing to do. particularly in a tool designed to simplify and streamline installs. The purpose (for pkg-get, at least), is NOT to offer *everything that everyone could possibly want to do*. It is to simplify installation, for the most common straightforward uses. If people want to do really really fancy non-standard stuff... they should be doing it themselves. Similar to the whole concept of offering binary packages in the first place. We try to offer a set of binary packages, choosing compile flags for the most common good. But if people want to get that 1% extra performance by using one-cpu-specific flags.... they should compile it themselves. That being said... there is a more general issue here about "what do we do when a package wants 'a database' installed?" I shall offer a separate email on that topic, so more people will pay attention :-) _______________________________________________ maintainers mailing list [email protected] https://lists.opencsw.org/mailman/listinfo/maintainers
