[nonconvergent]

Yeah, it just explains Dijkstra's and not very well. A big point is that Dijkstra's like many MST producing algorithms gives you A minimum spanning tree, not necessarily THE minimum spanning tree.

And why does the price of 2 drinks to the same vertex change? Shouldn't all edges on a vertex share the same cost if the edges are only weighted by the cost of 2 drinks? If the edges were a combination of drink and travel costs, maybe this makes more sense (though you still wouldn't see a $0 edge).

[teraflop]

> A big point is that Dijkstra's like many MST producing algorithms gives you A minimum spanning tree

In general, the spanning tree produced by Dijkstra's algorithm isn't a MST at all.