This is Rack middleware that provides logic for rate-limiting incoming HTTP requests to Rack applications. You can use Rack::Throttle with any Ruby web framework based on Rack, including with Ruby on Rails 3.0 and with Sinatra.

• http://github.com/datagraph/rack-throttle

• Throttles a Rack application by enforcing a minimum time interval between subsequent HTTP requests from a particular client, as well as by defining a maximum number of allowed HTTP requests per a given time period (hourly or daily).
• Compatible with any Rack application and any Rack-based framework.
• Stores rate-limiting counters in any key/value store implementation that responds to #[]/#[]= (like Ruby's hashes) or to #get/#set (like memcached or Redis).
• Compatible with the gdbm binding included in Ruby's standard library.
• Compatible with the memcached, memcache-client, memcache and redis gems.
• Compatible with Heroku's memcached add-on (currently available as a free beta service).

# config/application.rb
require 'rack/throttle'

class Application < Rails::Application
  config.middleware.use Rack::Throttle::Interval
end

#!/usr/bin/env ruby -rubygems
require 'sinatra'
require 'rack/throttle'

use Rack::Throttle::Interval

get('/hello') { "Hello, world!\n" }

#!/usr/bin/env rackup
require 'rack/throttle'

use Rack::Throttle::Interval

run lambda { |env| [200, {'Content-Type' => 'text/plain'}, "Hello, world!\n"] }

use Rack::Throttle::Interval, :min => 3.0

use Rack::Throttle::Hourly,   :max => 100

use Rack::Throttle::Daily,    :max => 1000

use Rack::Throttle::Daily,    :max => 1000  # requests
use Rack::Throttle::Hourly,   :max => 100   # requests
use Rack::Throttle::Interval, :min => 3.0   # seconds

require 'gdbm'

use Rack::Throttle::Interval, :cache => GDBM.new('tmp/throttle.db')

require 'memcached'

use Rack::Throttle::Interval, :cache => Memcached.new, :key_prefix => :throttle

require 'redis'

use Rack::Throttle::Interval, :cache => Redis.new, :key_prefix => :throttle

Rack::Throttle supports three built-in throttling strategies:

• Rack::Throttle::Interval: Throttles the application by enforcing a minimum interval (by default, 1 second) between subsequent HTTP requests.
• Rack::Throttle::Hourly: Throttles the application by defining a maximum number of allowed HTTP requests per hour (by default, 3,600 requests per 60 minutes, which works out to an average of 1 request per second).
• Rack::Throttle::Daily: Throttles the application by defining a maximum number of allowed HTTP requests per day (by default, 86,400 requests per 24 hours, which works out to an average of 1 request per second).

You can fully customize the implementation details of any of these strategies by simply subclassing one of the aforementioned default implementations. And, of course, should your application-specific requirements be significantly more complex than what we've provided for, you can also define entirely new kinds of throttling strategies by subclassing the Rack::Throttle::Limiter base class directly.

The rate-limiting counters stored and maintained by Rack::Throttle are keyed to unique HTTP clients.

By default, HTTP clients are uniquely identified by their IP address as returned by Rack::Request#ip. If you wish to instead use a more granular, application-specific identifier such as a session key or a user account name, you need only subclass a throttling strategy implementation and override the #client_identifier method.

When a client exceeds their rate limit, Rack::Throttle by default returns a "403 Forbidden" response with an associated "Rate Limit Exceeded" message in the response body.

An HTTP 403 response means that the server understood the request, but is refusing to respond to it and an accompanying message will explain why. This indicates an error on the client's part in exceeding the rate limits outlined in the acceptable use policy for the site, service, or API.

However, there exists a widespread practice of instead returning a "503 Service Unavailable" response when a client exceeds the set rate limits. This is technically dubious because it indicates an error on the server's part, which is certainly not the case with rate limiting - it was the client that committed the oops, not the server.

An HTTP 503 response would be correct in situations where the server was genuinely overloaded and couldn't handle more requests, but for rate limiting an HTTP 403 response is more appropriate. Nonetheless, if you think otherwise, Rack::Throttle does allow you to override the returned HTTP status code by passing in a :code => 503 option when constructing a Rack::Throttle::Limiter instance.

http://datagraph.rubyforge.org/rack-throttle/

• {Rack::Throttle}
  • {Rack::Throttle::Interval}
  • {Rack::Throttle::Daily}
  • {Rack::Throttle::Hourly}

• Rack (>= 1.0.0)

The recommended installation method is via RubyGems. To install the latest official release of the gem, do:

% [sudo] gem install rack-throttle

To get a local working copy of the development repository, do:

% git clone git://github.com/datagraph/rack-throttle.git

Alternatively, you can download the latest development version as a tarball as follows:

% wget http://github.com/datagraph/rack-throttle/tarball/master

• Arto Bendiken - http://ar.to/
• Brendon Murphy - http://www.techfreak.net/

Rack::Throttle is free and unencumbered public domain software. For more information, see http://unlicense.org/ or the accompanying UNLICENSE file.