NSCLDAQ version 12 introduced two format changes relative to NSCLDAQ version 11: The lack of a body header is denoted differently, Information is provided to trace back items whose body header source id was rewritten by the event builder back to their original source.
In NSCLDAQ-11, the ring item header is followed by either the size of a body header, if the item has a body header, or zero if it does not. A more consistent interpretation of the 32 bit word following the header was adopted in version 12; The lack of a body header is indicated by a 32 bit word containing sizeof(uint32_t). Thus skipping the body header, or lack of it, is handled consistently without a special case check for the lack of a body header.
It's important to know which readout program actually produced each item of data. In version 11, e.g. scaler data had a source id in its body header but the event builder overwrites that to the source id it was told to emit (in case there is hierarchical event building). Many ring items in NSCLDAQ version 12 have an added original source id field. The idea is that the originator can write its source id here as well as in the body header.
The original source id can then be used to determine which producer actually produced the item as it is immutable through the dataflow.
As with other concrete formats, the headers of the classes exported by this format are in a subdirectory of the includes installation directory. In our case the subdirectory is v12. Furthermore, all classes and definitions are qualified by the v12 namespace to prevent collision with classes in other format.