Interestingly enough, the published work [1] has two self-citations. So maybe there's something to that phrase, “the next work cannot be carried on without referring to previous work.”
I think Amazon would have benefited from better UI for particular parts of their platform and achieved better growth.
They're great for point of sale, terrible for product discovery as a direct result of the UI. That's fine for most products where you can research what you want to buy then purchase on Amazon anyway (because free 1/2-day shipping), but it means services like Prime Video are terrible to use compared to Hulu/Netflix.
Which is a shame, since I think their originals and content library is really solid. It's just impossible for you to discover it without a huge ad campaign.
Honestly? Because no software is free from bugs and there will always be a need for support. The difference between single sale and subscription can be life or death for a company, its products, and its support.
But if you don't feel that a subscription to a service adds value, please don't buy it and voice your concern to the developers. Especially if they're a small shop.
That said, subscriptions can be canceled. For example I needed Adobe Photoshop and Illustrator this month. Got to use those for a fraction of the cost I would have paid five years ago for each program and been stuck with a license. It's more flexible that way. But I understand it doesn't reflect every use case.
I wholeheartedly agree, but it's still not perfect. Introducing new revenue models into old industries where people are used to "owning" their software can be dicey.
A decent middle ground is permanent fallback versions. But then you still have the issues with feature creep.
>In Rust, does the standard way of implementing a graph structure still use a vector of nodes and indices for edges?
You'd probably want to do this whenever you implement a graph in a performant language, it's faster and safer (depending on the problem). Implementing edges as vectors of pointers is an exercise in memory leaks and data races, imo.
You can see an implementation with adjacency lists here:
imo you solve these problems by segregating the ownership logic and relationship logic of the data structure.
In other words the DAG would never be traversed to allocate or free memory, each vertex would be held by a vector or other container and the edges represented by an adjacency list/matrix.