I notice a subtle difference with trunk in the load order: Trunk does: 1) class def 2) compile new/changed methods 3) install new methods 4) remove methods (classes) 5) retry errors 6) notify 7) postload (initialize) Pharo does almost the same but delay method installation 3) after steps 4) and 5) 1) class def 2) compile new/changed methods 3) remove methods (classes) 4) retry errors 5) install new methods 6) notify 7) postload (initialize) Both solution might suffer from problems when changing a core package... The question is: is one better than the other? My opinion is that pharo has problems with refactorings: if I rename a core method, then it will be first removed before the new one be installed and that might prevent a package to load. What do you think ?