o2r ERC unpacker for ReproZip

[nuest]

This is just to collect some ideas of an offline discussion on a potential unpacker into ERC. It would sort of entail the Docker unpacker, but elicitate additional metadata from the user (with a nice CLI UI?) which ReproZip would normally not capture, such as author information, the main document and view file, etc. Before I even think about implementing such a thing, some questions:

@VickySteeves IMHO this is also a nice way to learn how the approaches and goals of o2r and ReproZip differ :-)

@remram44 We want to capture two aspects of the runtime environment, an image and a manifest, the latter is usually the Dockerfile for us. I'm wondering if you have an idea what kind of manifest ReproZip could produce... maybe this is the log of all traced files? Maybe it is the output of dpkg --list filtered by the libs which are traced (sounds tricky)?

[remram44]

ReproZip already does the tricky dpkg (and rpm) integration, capturing which DEB or RPM packages the traced files come from. But more could be done: Ruby/R/Python packages could be inferred in a similar way (which could be a base for your workspace-to-Docker process).

Note that ReproZip has support for plugins, that could collect additional information at packing time. It is only currently used to filter input files a bit, but for example, I envision looking for Git repositories at some point, so version information can be captured (only works if running from a source tree).