hypersycos / incrementalbackup Goto Github PK

In the instance of many small changes the current implementation will have lower read performance, particularly on hard drives. However, amalgamating files would be less robust in the face of power loss, or would require much more data to be written - decreasing performance in the write case. Perhaps a toggle would be best?

Being able to choose a handler based on metadata, magic numbers or MIME type would help alleviate this

A file must have all backups combined to compare itself to the new version. Heuristics such as checking last modified help, but do not fully alleviate the issue since the files with the most backups are the most likely to have been updated. Storing a full copy alongside the incremental format would provide O(1) reads as opposed to O(n). If done for all files this would increase the total file size by about 50% - but it could be done only for files which meet a certain number of changes.