[neilk]

I have not yet seen any application that would be improved by full HATEOAS. Fielding argues that you need it if you intend to be around for decades. Guess what, the Flickr API is verging on a decade and it's still doing fine.

Some things in the Flickr API could be nicer with more HATE. For instance, currently, the client needs to know how to turn a returned photo ID into a URL. Fielding would ask why the API doesn't just return URLs in the first place. (One answer: because that would be inanely repetitive.)

But that's mostly cosmetic. Yes, we expect the client to translate IDs to page URLs, but we also expect it to understand what the difference between a photo and a user is.

Nobody has explained to me how "Hypermedia" APIs are supposed to be traversable without a programmer embedding significant domain knowledge into the library anyway.