On Tue, May 4, 2010 at 1:26 AM, Reinhard Haller <reinhard.haller_at_interactive-net.de> wrote: > Am 03.05.2010 21:55, schrieb Garrett Cooper: >> Also, for services like cups, there could have per-application >> virtualized networking stacks > Hi Garret, > > one jail per application -- theoretically the best idea -- no conflict > due to the elimination of cross-dependencies. > Havig updated a server with 10 jails last week going thru 11 boring > mergemaster sessions I'm not convinced this a practicable way. > > Considering my problems with the update of all installed applications my > keypoints are: > 1) We have too much applications to manage ports, oftly you have to use > 2 different applications to do the job, so even forcing all applications > to compile/update doesn't eliminate the need to set up the update more > than once. > 2) Ports like db (40-50), python (2, 25, 26) need a proper handling by > the ports management. Over time I had installed 4 db versions; apr > doesn't compile with db >48. > 3) Configuration dependencies are not properly handled (Installing xorg > in a jail due to a unneeded configuration default is no fun). > > The goal of PBIs as Julian proposed is to simplify the automatic > generation of simple apps. > > To achieve this goal we get another ports management application and > hope it handles also the non trivial tasks of the non simple apps. > > If the PBIs come with all libraries and resources we get even more > problems with multiple db installations not less. > > Are configuration dependencies (exim with or without ldap) addressed > with the PBI format? > > I believe we need a more precise way to express the dependencies between > the ports. I'm not going to feign knowing what's going on completely in this regard; if we were pointed to the software spec(s) for PBIs and the tools, it would probably make analysis easier. Thanks, -GarrettReceived on Tue May 04 2010 - 15:16:26 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:40:03 UTC