DAP4: DAP4 Using Multi-part Mime
From OPeNDAP Documentation
⧼opendap2-jumptonavigation⧽
Background
As a result of a comment in the OPULS/NOAA telecon, I think we should reconsider using multi-part mime in a limited fashion in our serialization. Basically I propose to resurrect James' original proposal (see DAP 4.0 Design).
Proposal
As with the original proposal, the multipart-mime has two parts. The first is for the DMR and the second is for the whole, chunked serialized data. The general form would look something like this.
Content-Type: multipart/related; \ type="application/vnd.org.opendap.dap4.dataset-metadata+xml"; \ start="<<start id>>"; boundary="<<boundary>>" --<<boundary>> Content-Type: text/xml; charset=UTF-8 Content-Id: <<start id>> Content-Length: ddddd <<blank line>> <<DMR text>> --<<boundary>> Content-Type: application/octet-stream Content-Encoding: big-endian Content-Length: -1 <<blank-line> <<chunked representation of the serialized data>> --<<boundary>>
Notes:
- Every header line should be terminated with CRLF. The <<blankline>> is also terminated with CRLF.
- The above representation assumes that the DMR is not in chunked format.
- The Content-Length value for the DMR counts the number of UTF-8 characters from the end of the blank line to the initial "--" of the boundary line.
- The chunked representation is still in our existing chunked format.
- I do not know if the start-id is required by the multipart-mime format. If not, then we should remove it.
- I am not sure about the "type" param of the "Content-Type" line at the beginning.
Dennis Heimbigner Initial draft 3/19/2013