The first step to getting smart about setting up appropriate segment grammars for HL7 messages is to realize how non-standard the standard is. Since no two vendors implement HL7 the same way, HL7 is often referred to as the "non-standard standard". Setting up an HL7 interface is not a matter of selecting the standard HL7 messages from the libraries as shipped, but involves looking at the data you find and designing your grammars accordingly.

In practice it doesn't matter whether a vendor claims to have implemented 2.3.1 or 2.1 or the 2.4 version of HL7. The only thing that matters is what they actually implemented. The specification most vendors provide for their HL7 interfaces is typically not accurate. If you really want to see what was implemented you'll need to start working with the real data you acquired.