> > i am not too familiar with the sync mechanism... but looking just > quickly at the code it seems that it persists the sync of each user/group. > that looks reasonable to me (even if it comes with some Root.commit() > overhead) as it allows to specifically retry or revert the changes > made during the sync of one single account. that's the defensive > way of minimising the unexpected behaviour for the consumer in case > of failure. > > on the other hand you might argue that the default should be success > and that in those cases less frequent calls to persist the changes > might work as well and was probably faster. mmm interesting, yeah i might argue that :-) and that we could probably in case of an error “replay” user/group sync persisting one by one, from the last successful save. Will have a look.
Thanks a lot, Nicolas
