I watched this talk, and I'm agreed that there's no need in AOP technics 
provided by DI - because Ruby is dynamic language and has 
metaprogramming capabilities that are way more powerful and handy than 
DI AOP.

But I disagreed with the conclusion out of this fact - that there's no 
need for DI in Ruby. Because AOP is only one aspect of DI, and although 
You don't need AOP, You may benefit from others capabilities.

No matter how Your code looks like - lots of Lego building blocks (Java) 
or free-design with plasticine, You has to solve following problems:

- where the component's code is located.
- in what order should it be loaded.
- what configs does the component needs to be properly initialized.
- where those configs are stored.
- how to change configs in different environments (dev/test/prod 
database settings, routes, ...).
- where are dependencies for component and how they should be 
initialized.
- how to replace some components with custom implementation (without 
examining its code and monkey-patching).
- how to assembly parts of application for specs/tests.
- how to restore state after each spec/test (isolate it from each 
other).
- how to control life-cycle of dynamically created components.
- connecting components to assemble an application.

DI simplifies these tasks and some of them solves completely 
automatically.

And if You think that You don't need any structure and want complete 
freedom - why do You use Rake?
It limits Your freedom to work with the Ruby plasticine and introduce 
extra abstraction, so why do You use it?

-- 
Posted via http://www.ruby-forum.com/.