diff --git a/web/spec/testing_oauth.txt b/web/spec/testing_oauth.txt
new file mode 100644
index 000000000..ed79066a7
--- /dev/null
+++ b/web/spec/testing_oauth.txt
@@ -0,0 +1,19 @@
+For access to the youtube and google APIs, we need an access_token
+
+To obtain an access token, one must actually log into google using a browser running javascript.  This redirects to the URL specified, as long as it is specified in the oauth configuration.
+
+Getting an access token for the purposes of automated testing is tricky, but possible using Capybara with a javascript-enabled driver.  (Note, web/spec/support/utilities.rb utilizes the JK youtube client to perform the intricate bits):
+
+1) Obtain the login URL.  It's ugly, but we can get it from the YouTubeClient.  It contains the callback URL, as well as a "hint" that will fill in the username for us.
+2) Start a web server on an enabled callback server, such as localhost:3000
+3) Obtain the URL using a known test user
+4) Visit the URL in a capybara test
+4a) Fill in password with the right value
+4b) Click the login button
+4c) The approve page should load.  Wait for the approve button to be enabled.  This is usually a second or two after the page loads, but not immediately.  
+4d) Click the approve button
+5) After google approves, some javascript will redirect to our test web server, which contains a code.  This is not the access_token, but a one-time code that can be exchanged for an access_token, again POSTing to google's auth server.  You can see it in gory detail in YouTubeClient.exchange_for_token.  
+6) If all goes well, the test web server will call back the invoker with a real access token. 
+7) For testing purposes, stick the access token in the user.user_authorizations table for the user for which we are testing.
+
+