Pildream wrote: as far as i know when a rollback happens everyone should return to the state they were before, so nothing should be missing, right ?
Where information integrity is 100% important, yes, you can expect this to be done. However, very serious server time, storage, and/or other resources are dedicated to doing this. For the rest of us, we lose data when we have to roll back to a prior "save." The amount of data saved can be described along a curve based on the time and money spent on pushing towards 100% redundancy. I wouldn't call it exponential, but it grows at a rate faster than linear*.
Of course, in an ideal world, we wouldn't have server crashes. But this is a beta for these reasons. Reliability can't be guaranteed ever, even in the most redundant of systems, but this is a work in progress with major changes being made fairly frequently, so reliability is something we can't even think about.
Note: JC made a post here about his decision:
viewtopic.php?f=25&t=58&start=990#p187887*There have been technologies created to help lower these costs, but they still require server performance that may not be available to smaller companies, and then it takes time to rebuild the database from the logs generated... sometimes a good bit of time (hours or days depending on the system). Even then, they are only as reliable as the information fed to them, so if the information is bad, the same crash is going to occur and you still lose subsequent work.