I find these for Presentation Data Value:
0 | 0 | 0 | 0 | 0 | 0 | last-or-not | command-or-dataset | **Some Message**
But I couldn't find Some Message part. I suppose this part include C-Find, C-Get etc. How can I know this structure?
Where do you have this from? In fact it is a bit different.
Your example should read
0 | 0 | 0 | 0 | 0 | 0 | last-or-not | command-or-data*set* | **Some Message**
So the "command-or-dataset" flag indicates whether the following bytes are encoding a command (as defined in PS3.7) or a dataset as defined in PS3.3 or 3.4 respectively).
E.g. for DICOM Queries, there is a C-FIND command defined in PS3.7, Chapter 9.1.2.1. In C-FIND, the query criteria are part of the command ("Identifier") in table 9.1-2. How the identifier is formed and all its semantics is subject of the Query/Retrieve Service Class as defined in PS3.4, C.4.1.
For transferring objects, there is a C-STORE command, also defined in PS3.7 (chapter 9.1.1.1). The Data-Set is also a part of the C-STORE command, and its contents depend on the type of data (SOP Class). This is referred to as an Information Object Definition (IOD) and defined in PS3.3. The protocol for Strorage is also defined in PS3.4 (Annex B) However, the length limitation of the PDV will only allow to have the whole object encoded in a single PDV and needs to be split. For the following PDVs, no command set will be present but only a fragment of the dataset. In this case, the "command-or-dataset" bit must be set to 0.
I hope I could make it a bit clear. It is a bit difficult in the beginning of learning DICOM to know all the terms and the interrelationships.
Encoding
Logically command- and dataset are encoded in the same way. The data dictionary (Part 6) is a complete list of all possible attributes and the major difference between command- and data set attributes is that command attributes are having the group number 0 while data set attributes have "any even number but 0".
For each attribute, the data dictionary gives you the Value Representation (VR) which needs to be considered for encoding the value. E.g. "PN" for patient name "UI" for Unique Identifier and so forth. The VRs are defined in PS3.5, Chapter 6.2. The encoding of attributes is then
group | element | (VR) | length (always even) | value
How this is transformed to the binary level depends on the Transfer Syntax (TS) that was agreed for the service during association negotiation. For this reason "VR" is enclosed in brackets above - it depends whether it is an implicit or explicit TS if this must/must not be present.
There are some more things to consider (endianess, sequence encoding) when encoding code sets or datasets in binary form. Basically everything about it is described in various chapters in PS3.5