We're not on Ruby 1.9.2–we're on Ruby 1.8.7 (actually, REE).

Try this:

1. generate scaffold title:string body:text (sqlite3)
2. run server in production mode
3. ab -n 1000 http://localhost:3000/posts (on 20 records)

I've got about 50 req/s for Rails 2 and only 35 req/s for the same page, which is strange after all that buzz about performance optimizations in Rails 3. It seems all improvements was ruby 1.8.7 only (haven't checked if Rails 3 is really faster on 1.8.7 though).

Also in development environment performance degradation comparing to Rails 2 is even worse :(

Anybody have similar results?

template_for = Hash.new{|hash, object| hash.update object.class => compute_template_for(object.class)}

…

@collection.each{|object| template_for[object]….

that way you don't have to do any up front template selection and worst case == best case == one lookup

There is nothing stopping you from using it right now though. Rails master has been stable for some time now. All tests are passing. I've been able to generate and run an application off Rails 3 for some time now.

So rather than saying "RoR3 is not ready for serious usage", say exactly why it isn't ready so developers might be able to address the issues. Statements like that without explaination don't help anyone.

* Increasing Performance and reducing overhead (this post)
* Separation of responsibility and Dependency declaration (previous post)
* ARel integration with ActiveRecord (no blog posts I know of, but a lot of commits for it recently)
* Rails router DSL improved – http://rizwanreza.com/2009/12/20/revamped-routes-… and http://yehudakatz.com/2009/12/26/the-rails-3-rout…
* Gem bundler (allowing references to Rubygems to be removed) – http://yehudakatz.com/2009/11/03/using-the-new-ge…
* Major cleanup / refactoring of internals (particularly around dispatching/respond_to)
* many many more….

As for timelines, while there isn't a specific one available, mainly because with project like this, it is very hard to accuratly predict them, DHH has estimated a beta release by end of January: http://twitter.com/dhh/statuses/7208225785

From what I have seen regarding plugins, Rails 3 will maintain backward compatability, but includes new features such as off loading the rake task loading to the plugin, and adding support for initializers within the plugin. As an example of this, ActiveRecord is now treated as a plugin of Railties (which should make swapping out the ORM a lot less painful). Example: http://github.com/rails/rails/blob/master/activer… . These added abilities should make development via gems a lot easier.

As it has been said many times before, Rails 3 API will be almost entirely backwards compatible. There are some changes in public API (like router changes and new responds format), but huge amount of work was done *inside* Rails.

It implies to things:

1. You don't have to wait for Rails 3, most of the knowledge learned for rails 2 will be still valid.

2. Almost all the plugins will be broken, so if you start with Rails 3 try to not use big plugins.

So the short answer:

If you are newbie start with rails 2.

I've started writing application in Rails 3 and it basically works. I've managed to rewrite some gems using new Rails 3 modules (ActiveSupport::Callbacks, ActiveModel::Validations and so on) and it's well structured easy to understand code.

If you feel that you know very little, at least *try* to find any information. Merb was also not so well documented and you had to figure out many things from code.