The planner behind a mikrolet route service traded optimal paths for rules that were easy to change.
Depth-first search has a mixed reputation in routing. It does not guarantee the shortest path, and on a dense network it can wander down long branches before finding anything useful. When a team picks it anyway for a public transport trip planner, the reason is worth understanding.
In the Malang mikrolet study, the authors are upfront about the trade. They chose depth-first search because it is relatively simple to modify, and the routing rules for paratransit were still being worked out. Producing correct plans mattered more, at that stage, than shaving milliseconds.
The other half of the design is where the plans live. Rather than baking routing into a phone app, the team exposed it as a REST web service in a location-based style. A request carries an origin and a destination. A response carries candidate journeys.
Functional testing confirmed the plans came out as expected. That is a modest but real milestone, because a route planner that returns plausible but wrong itineraries is worse than no planner at all.
The same choice that made the planner easy to modify makes it expensive to scale. As routes and possible plans multiply, the search slows down, and the authors flag optimization as the next step. Depth-first search was a fair call for a prototype. The paper does not pretend it will carry the system forever.
For anyone sitting on legacy transport data, the order of operations is the lesson. Ship something correct on a simple foundation, learn what riders actually need, and optimize once you know which problem you are solving.
Kharisma and Pinandito (2019). Prototype of REST Location-based Service for Trip Planning by Paratransit (Mikrolet) in Malang City. Journal of Information Technology and Computer Science.
