I've read a similar question and Raymond Chan's blog post, but still have a question around using a shortened GUID as an id. I understand that GUIDs are unique, but substrings of GUIDs are not.
I am building a site that supplies the user with a reference number for an order. The client does not want to use an auto-incremented Int, as this will expose the total number of orders. They also consider GUIDs too long and difficult to repeat over the phone.
However, as we will only have one web server, it has been suggested that we can use the non-MAC address part of a GUID, to give a shorter, unique identifier.
My understanding is that the current GUID algorithm does not allow for this. Is this correct? Can GUIDs created on the same machine be shortened yet still unique?