A benchmark of the speed of two options for backporting Ruby 1.9's public_send method to 1.8.7. The first one wins by far.

public_send_benchmark.rb

require 'benchmark'

class Object
  def public_send_public_method_defined(method, *args, &block)
    singleton_class = (class << self; self; end)

    if singleton_class.public_method_defined?(method)
      send(method, *args, &block)
    else
      raise # this doesn't matter, as it's not the path we're benchmarking
    end
  end

  def public_send_methods(method, *args, &block)
    if methods.include?(method.to_s)
      send(method, *args, &block)
    else
      raise # this doesn't matter, as it's not the path we're benchmarking
    end
  end
end

class Foo
  def bar
  end
end

n = 100_000

Benchmark.bmbm(20) do |b|
  b.report('public_method_defined?') do
    n.times { Foo.new.public_send_public_method_defined(:bar) }
  end

  b.report('methods') do
    n.times { Foo.new.public_send_methods(:bar) }
  end
end

results.txt

$ ruby public_send_benchmark.rb 
Rehearsal ----------------------------------------------------------
public_method_defined?   0.530000   0.000000   0.530000 (  0.533415)
methods                  2.820000   0.000000   2.820000 (  2.833929)
------------------------------------------------- total: 3.350000sec

                             user     system      total        real
public_method_defined?   0.530000   0.000000   0.530000 (  0.529561)
methods                  3.140000   0.000000   3.140000 (  3.155589)