I think it was Chandler that wrote about how genre is both a super useful tool for grouping things and nearly impossible to define in a way everyone can agree on.

Hierarchical structure I think is more or less dead at any degree of complexity, you see this across so many different systems that are trying to make browsing and searching big collections of data intuitive, trying to move away from creating “folder” like structures.

I remember listening to a talk on component-based systems at a GDC back in the early 2000’s and back then it blew my mind. Such a good way to solve the multiple-inheritance problem in software development, where OOP (very good in so many ways) had problems when it came to sharing properties, you either had massively growing base classes or the multiple inheritance look-up problem. Now more modern languages use interface fulfilment, traits, mix-ins and many different techniques to resolve this issue. However I still really favour the component-based approach, which was mooted way back when (and used to great effect in many game engines, like Unity for example). Taking that same principle and applying it here I think is how I would go personally.

That is to say, rather than try to categorise items, allow items to have other common component associated with them to help search and categorise them. At the most basic level, this can just be tagging, a little more sophisticated is allowing tags to have sub-tags which can be flagged on and off (e.g. Tag Historical, sub tags might include “WWI”, “American Civil War”) etc… You can go a lot more complex than that, but already that gives a lot of power for filtering and searching.

You could say, why allow sub-tags at all? Why not just have everything as a flat list of tags. Well one thing is helping auto-fill on tags, if I pick Historical, it makes sense to suggest sub-tags associated with Historical, as opposed to anything else. Though a lot of modern systems would probably go with suggesting along the “people who tag X, also tend to tag Y”.

Just my thoughts though. Whilst my technical background of recent years has involved some big-scale data and processing, my main technical focus is still on getting really pretty 3D graphics around the world, my colleagues in the data-engineering team might roll their eyes at my game-engine one-track-mind 🙂

I’m also sure the team have been considering all sorts of problems around DB indices, scaling, redundancy and so on, and probably already thrashed out and threw away tons of better ideas than anything I am suggesting here to arrive at what worked best!