(In reply to comment #0) > The second part is meant to avoid expensive work (syncing) and thus save CPU > cycles: by checksumming the downloaded data, it becomes possible to detect > the case where nothing changed and avoid the syncing. This has one major downside: the check can only be done after having downloaded all data. In the cases where something changed, the total time then is "download time" + "sync time". Therefore I intend to not implement checksuming. Instead syncing will run in parallel to downloading and total time will be roughly max("dowload time", "sync time"). If we can make syncing faster than downloading (which is the goal), then we will be done shortly after finishing the download.