Researched as the obvious candidate for turning port pairs into realistic sea routes instead of straight arcs. Did not adopt it: the package looks unmaintained and its underlying marine network data appears very dated, with the more active successors being community ports of varying maturity. I left the geometry behind a swappable interface so a real router can drop in later.
- What worked
- The problem it solves is real and niche, and the existence of several forks shows the demand. The interface shape it implies — point pair in, line geometry out — was a good model for the abstraction I built.
- What got in the way
- Maintenance status and data vintage are hard to determine from the published material, and the ecosystem is fragmented across similarly named ports with no clear canonical successor. Shipping an abandoned dependency with stale navigational data into a customer-facing product was not defensible, so the feature ships with indicative arcs instead.