iwarshak / secure_actions

***************************
* Do not use this plugin! *
***************************
Updated 5/31/2008

This plugin is deprecated. I would suggest using this plugin instead:

http://github.com/iwarshak/ssl_requirement

I have taken the approach of making landing pages (such as the login form, credit card form, etc) be SSL required, and 
then using *_path type URLs and not *_url type URLs. This seems to work just as well, and is a lot less complex.
 

SecureActions
==============

This plugin allows you to specify actions that must be run under ssl. If they are accessed 
without ssl they will be redirected. This is similar to ssl_requirement 
(http://dev.rubyonrails.org/browser/plugins/ssl_requirement/). In addition, if a link is generated
to a secure action (using url_for, link_to, etc) that link will be an http:// link.

The benefit to this is: If you are only relying on the http to https redirection for security then 
by the time you are redirecting the user to use https, data has already been transmitted insecurely.

By declaring which actions you want to be "secure" than any links to those actions will have https:// links
and if, for some reason someone tries to access that page with http:// then they are going to be redirected 
back to https://

Credit to DHH and his ssl_requirement plugin
Also Duane Johnson and the folks on this mailing list thread: 
http://thread.gmane.org/gmane.comp.lang.ruby.rails/13488/focus=13493

Contact me: iwarshak@stripey.net or http://www.ianwarshak.com

Usage
===============

environments/production.rb (or whatever environment you want SSL enabled)
USE_SSL=true


class MyController < ActionController::Base
  include SecureActions
  require_ssl :index, :secure_form
end


Notes
===============

This plugin generates overrides default_url_options to *always* generate full urls instead of relative
urls. Otherwise we would never be able to switch modes from http -> https. So if are linking to a
secure action with link_to, the link you get is an https:// link. 

The one issue that I have found with this approach is that page caching relies on url_for to generate 
the location on the filesystem for the cached pages. Normally the cache_page method would call
url_for(:controller => "foo", :action => "index") and get back /foo/index. It would write the response
to CACHE_ROOT_DIR/foo/index.html
CACHE_ROOT/
->foo/
-->index.html


Since we are forcing FULL urls to be returned from url_for, this would cause ruby to try to write the
caches page to http://foo/index.html. On a unix system you end up with something like this.
CACHE_ROOThttp:
-> mydomain.com/
--> foo/
---> index.html

Obviously if your webserver (httpd, nginx) is going to have a hard time figuring this out.

The solution was to add another option in url_for which is an override for only_path.
Remember, we have set only_path to *always* return false and force the full http://host.com/controller/action
style url.

So if :override_only_path is set, we allow only_path to be set to true. Then I overrode
the page caching methods to call url_for with this option and we get sane paths