The existing Wikipedia prototypes using serviceworkers are generally caching every article the reader browses by default. We can do easily, but does it make the interaction more confusing?

Our finding specifically talks about people downloading/saving content to read/share later, so it's not clear that this really hits what they want. We could include both, but that might make the UI more confusing.

Another consideration is that with limited storage, we would want to prioritize reader-selected content over the history.