Mask oAuth API key and token for pure client-side technologies

Section 4.6 of RFC 5849, which defines OAuth 1, states that the consumer secret was never intended for use by desktop consumers, despite Twitter's use thereof in practice. As Nelson Elhage pointed out in "Dear Twitter", Twitter can and does terminate consumer keys of desktop clients, provided that the client isn't too big to fail. But there are two workarounds to allow use of OAuth 1 in a desktop or mobile application.

One way is to proxy the entire Twitter protocol through a server that you operate. This way, the consumer secret stays on your server. This is the workaround recommended by Dick Hardt, editor of the OAuth 1 spec. This workaround does not address the cost of operating this server.

The other way, as suggested in a post by Raffi Krikorian to the Twitter development talk Google group and a post by Chris Steipp to a Wikipedia mailing list, is to "have each user register their copy of your desktop application as its own consumer." Then the user would copy and paste the newly registered consumer key and consumer secret into your application. The manual for your application would then need to include detailed instructions on how to register a new application on Twitter's developer site. This official limitation has a few practical problems:

• Your client will face a usability disadvantage compared to well-known proprietary clients.
• The form to create a new app doesn't appear to offer a way to pre-populate the required fields. This means you will have to update the registration walkthrough in your manual whenever Twitter changes the procedure for registering an app.
• The developer agreement requires users to be of legal age to enter a binding contract. This means a user of your application aged 13 to 17 must have a parent accept the agreement on the user's behalf.
• Twitter's Developer Policy prohibits mass-registering applications and "name squatting", which it defines as "submitting multiple applications with the same function under different names." I am unaware of any precedent as to whether Twitter has treated unrelated users who have registered separate copies of one application as "name squatters".

Your performance gain are going to come from Scaling up, (if you need to). Imaging you have a whole lot of users Option 2 would cause server side bottle neck much sooner than option 1.

Small scale where the server would have no problem with the load, the performance gains would be negligible.