If we do attempt a large-scale import of an existing road dataset, especially a snapshot-in-time dataset, we should improve upon the process used in previous road imports such as in Santa Clara County so that this fledgling community isn’t left with more cleanup work than it can swallow. We also urgently need to resolve some best practices questions, particularly how to model changes over time to attributes besides route designations.
For the time being, I would favor making it as easy as possible to use NTAD as a comparison layer or importing it piecemeal without directly importing it in bulk. For our road coverage to be a serious resource, it needs to be granular down to individual alignments and realignments, not just putting one date on an entire cross-country route. In many cases, this will require researching a road more intensively than what the Wikipedia article provides. A bulk import may obscure opportunities for deeper historical coverage.
Lots of stellar work has gone into the road coverage we do have. I think much of it is motivated by the desire to light up roads on the map, but some of this motivation would be swamped by a bulk import.