Hi, rdiff-backup has currently its own file formats, which are far from being standard, meaning a lot of custom code to handle these formats, respectively a lot of different files.
Middle term I'd like to move to a more standard format for code efficiency reasons: if the code is in a library, I don't need to write and maintain it, and newcomers don't need to learn the format. My opinion is almost made on which format I'd like to use but I want to make sure I don't miss arguments and hence open the discussion before it's too late. On the basis that we have 4 potential formats to consider (XML, JSON, YAML or TOML), I've put together the following criteria: - human readable, standardized, well-maintained Python library available, the obvious... (all) - allows for comments, useful for debugging and testing purposes (JSON out) - as little verbose as possible, easy to create/fix manually (XML out, JSON ok-ish) - complete language, i.e. allows for dicts, lists, lists of dicts, etc., up to a full-blown serialization language [1] (TOML out) - streamable, i.e. doesn't need a closing element to be valid, because our current formats are like this, and it fits better the iterable-approach used throughout the code (JSON and XML out) I was almost finished with this list, as I discovered [2] which in many more words and examples are saying quite the same. Side note: even if the arguments are good, I'm not planning to use strictyaml, a library not even packaged for Fedora. I'd prefer to use the safe_load function of pyyaml, or whatever fits the needs (YAML 1.2 [4] might even get support in the near future [5]). As written, my point is about simplifying the code, no immediate decision required, but I prefer to have this discussion before it's too late to have it... Thanks, Eric [1] https://en.wikipedia.org/wiki/Comparison_of_data_serialization_formats [2] https://hitchdev.com/strictyaml/why-not/ [3] https://pyyaml.org/ [4] https://yaml.org/spec/1.2/spec.html [5] https://github.com/yaml/pyyaml/issues/116
