I'm aware that the DocumentDB team is working on an update to the service that will allow access & querying of documents, collections and databases with the user given ID versus having to use the service generated SelfLink. I am very enthusiastic about this update and based on conversations at //Build/ & Ignite - it sounds like it may roll out in the next 4-6 weeks.
My question is: once this is out, will there be a performance difference between using SelfLinks and ID's? Currently my application is storing the SelfLinks so I have fast access to them for my queries, and avoid the extra hop to get the SelfLinks. If I can abandon this overhead and use my own ID's without any performance hit I will gladly do so. If the service is merely doing this extra hop on my behalf to retrieve the SelfLink, then I will keep my current architecture to get the performance increase per call.
