We've discussed both proposals on todays call and based on the information available to the participants (github votes, ML comments) decided to move forward with PR #60. Across all the comments we collected, there seem to be minor issues with each approach, but overall #60 seem to be the preferred one. Please don't see this a negative if #59 was your preference, but more as an invitation to iterate with us on the code base (read: your PR's are welcome).