Hacker News new | ask | show | jobs
by surrealvortex 3713 days ago
Or maybe the transportation resource Amazon had to use to ship later in the day had some cancelations?

It's impossible to build a distributed system that has fully consistent results. That is usually the reason for inconsistencies between the detail page and checkout page promises.

1 comments

I guess you've indirectly answered my question - Amazon has its own reality distortion field. Are you really not aware that every other merchant and shipper has specific cutoff times? It's based on truck schedule rather than link capacity - the essential measure is weight, which primarily necessitates more fuel.

The only "distributed system" in question is Amazon's own computer system. In which case no, it does not really seem that difficult to distribute a basically static scalar field.

Um, if someone else cancels, they free up the resource (volume or weight or number of packages). This allows some other package to use the resource. This essentially makes it non-static, since it depends on the number of shipments that will consume that specific resource.

Your reply seems to suggest that you can get a truck to carry infinite capacity just by using more fuel. Not sure if you meant something else.

Space/weight limits of trucks are not the critical factor - sure, they're "infinite" if you insist on thinking in the wrong paradigm.

This isn't the Internet where links have a bounded capacity and zero marginal cost. The trucks/planes have their schedule - more packages mean more fuel is used, not exhaustion of discrete slots. Delivery services do not make money by queuing packages.

What? Trucks do have a bounded capacity. A truck cannot carry more than x kgs of packages or y liters of volume. Are you arguing against that?

If Amazon had asked for a single truck from UPS at 5 PM from warehouse A (the number of trucks ordered is a business decision), it can only allocate x kgs of packages or y liters of volume to that resource.

Edit: Just to clarify, the number of trucks may often be decided waaay in advance.

I'm arguing that the capacity of the truck generally does not matter. A toilet bowl has a bounded capacity too. But by the time you're thinking about it, you have a different sort of problem.

The cutoff windows they do give are too long to support your theory (what if demand then spikes?), unless Amazon also retracts those windows (which would also be customer hostile). I'm not saying the constraints you describe are impossible, just highly unlikely based on how every other merchant/shipper works. The tiny gain from optimally packing trucks does not seem to outweigh the additional complexity required to do so.