Time, and the evolution of science it brings, is an amazing thing. Once, developing a nuclear bomb required gathering a super team of scientists, closed off in a campus inventing bleeding edge chemistry and physics. Today, the workings of nuclear bombs are considered trivial by scientists in the area.
Dijkstra's algorithm is, by today's standards, pretty trivial. It's a breadth first search in a graph. Simple enough to be completely described in a single sentence. I'm not negating the brilliance of the scientist, I'm merely stating software has evolved in the last decades.
If the algorithm is simple enough to be described in one sentence, a software engineer, self taught or not, better be able to get there, when asked for a solution to the problem. Again, knowing it by name is irrelevant, but when asked for a shortest-path algorithm for a graph, any half-decent developer should be able to reach a reasonable approximation.