Implement data package download
#1 Updated by ben leinfelder almost 8 years ago
- Assignee set to ben leinfelder
For reference, there is this draft architecture spec: http://mule1.dataone.org/ArchitectureDocs-current/design/DataPackage.html#package-serialization-using-bagit
#2 Updated by ben leinfelder almost 8 years ago
- Status changed from New to In Progress
Aside from adding rendered metadata to the package, this is ready for group review.
Outstanding issues that I see:
-the entries in the package are named randomly (temp file creation) but the pid-mapping file tells users which file represents which pid.
-we are using lots of temp files. all but one are removed immediately, but the last bagit.zip file is streamed by the MN implementation so we don't really know when it is safe delete the file. It is marked for deletion "on exit" - but that could be weeks if not months if Tomcat is not restarted. Alternative is to hold the zip file in memory and stream from those bytes...but that seems even worse than having temp files build up.