This article digs on the flattened dependency model like it's a huge weakness of Bower, but this is the _entire point_ of Bower.

You include two models that both depend on two different versions of jQuery. NPM wouldn't complain at all, Browserify would happily package both dependencies up, and now your users are downloading jQuery twice to view your page.

Bower, conversely, raises a red flag right away, and lets you determine what should be done here in the event that the package definition doesn't contain enough info to sort this out automatically. Its goal is to make your client code as minimal as possible.

Feature, not bug.

Npm/browserify will load only one copy of jQuery if the semvers match. The only "problem" with that is that if you get duplicates, npm doesn't report these out of the box. It's rather easy to write a script to detect these (if one doesn't exist already).

I think Bower's model of dependencies is better than NPM's. NPM's trees are huge and ugly, when Bower's flat structure is more suitable for frontend apps. For example, if I will use npm for all libs I used to use in frontend web app, npm will install 2-3 versions of Angular (because so plugins still have Angular 1.2 in dependencies, some want 1.3 and I use 1.4). And what will we do with all these versions? Nothing, we can run only 1 Angular simultaneously.

I've never actually encountered this particular problem with bower. My main problem with it is that it's a pain to get it through corporate proxies. And I have the same problem with npm.

why do you need one library to load another? Am I missing something here?

I'm trying to move away from using Bower and just stick with NPM, but that's purely to simplify my project setup. I'd rather have one package manager managing all my dependencies.