Code review comment for lp:~javier.collado/utah/bug1087620

I don't think the lack of reproduction of symbolic links is an issue for our current purposes. I mainly put that in before to try to save space.

The previous validation looked for anything in stderr when extracting the files. It looks like we're discarding that now. Should we be looking for errors, in stderr or in the return code of the extraction process? I think it might be nice in general to raise an exception if a file fails to extract.

At some point, a similar overhaul of the dump function (to return the contents of the file rather than a subprocess) would probably be nice as well, but that's not urgent.