Morpho Refactoring Discussion
---------------------------------------------
Participants: Matt Jones, Ben Leinfelder, Jing Tao, Dave Vieglais

Design diagrams
http://chico1.dyndns.org/morpho/

Goal: Users should be able to use Morpho with Metacat repositories that are not registered DataONE nodes


Here is a proposal for implementing Morpho in the different phases:

Phase 1:  Work with multiple MNs, but choose one as active MN for create(). Morpho still uses Metacat API to search
               User selects 'Active' Member node from dropdown list in settings; list generated from Node service; or enter baseurl
                 If MN is metacat, use metacat API to search
                        ** Metacat search API only returns metacat docids (scope.docid.rev) not the DataONE PIDs
                        ** Issue: how to get ahold of the PID rather than the localID so get()/create()/update() can be called
                 If MN is not metacat, disable network search, only allow local open
                 For create
                       Create on the active MN, set authMN to the active node id
                 For update/delete                        
                    If a user has rights to call update()/delete() on the AUthoritative node, and a write permission on an object, they can update/delete on the AuthNode
                          -- 3 checks have to pass: userHasWrite on object, user has execute on service, and authMN is available for object
                           If authMN is not available, don't do update()

Phase 2: Use dataone API to search
             Some issues need to be addressed:
                 Will Morpho search CN, MN or both? If search MN, we need an optional search API for MN.
                 If we only search CN, we need to deal with the variable sync lag times on CNs
                 Need a search mechanism to include metadata fields in CN/MN
                 
Issues to be addressed:
-----------------------------------
 * restructure identifiers in Morpho to be opaque, support multiple formats (DOI, UUID, etc)
   * Can use MN.generateIdentifier and MN.reserveIdentifier
   * Need to implement these methods on Metacat
 * Need to restructure local storage to not depend on  the identifier format for directory layout
   * Need upgrade mechanism to transform current storage structure to new structure
 * Need mechanism to track revision history based on obsoletes/obsoleted by
 * Need mechanism to track package membership
 * Need ability to generate ORE packages
 * Local vs. Network versioning (new identifiers for local objects?)
   * 'Offline' mode -- how to generate IDs
 * Need to choose 1 (or more) MNs to use for CRUD ops, etc
   * Present list to user from results of listNodes() call
     * Once user chooses 'Active' or 'Working' node, all create() ops go there
     * In first iteration, all update() ops go there too
     * In later passes, potentially allow update() ops to go to other nodes too
       * Need to consider that the Authoritative node may not want other nodes to obsolete theiir content
   * Decide how to choose a CN
 * Given a CN search result, which MN is used to 'open' an object, and where is updated
 * Morpho should check service access control lists to see if user has service access before attempting a create() or update(), etc.
   * This is in addition to the per-object ACLs
 * MN terminology from the user's perspective
   * 'Working' Member Node (Maybe 'Active' Member node)
   * 'Authoritative' Member Node
 * Need to be able to turn on MN services on Metacat without registering with DataONE (standalone mode)
 * File storage issues
   * How can we let user's choose where their data are stored?
   * How can we let users have more than one directory for data storage?
   * How should files be named on disk?  Don't use PID because can have illegal characters
     * Need file naming convention
     * So need mapping between PID and filename
     * simple delimited text file at root of the data dir?  ignored by hashing scripts, used to map ID/filename
   * Consider using JCS to control file storage

* {filename (Morpho creates it), PID/DOCID}

Query stuff
-----------------
 * Need search mechanism to include metadata fields in CN
   * CN search should include user-defined list of fields
     * Need to determine how to express that as a result set structure
       * SOLR syntax for resultset?
       * opensearch syntax for resultset?
 * What should Morpho search?
   * Production CNs?  Configurable to other CNs?
   * Member nodes?  Like metacat servers?  Need a MN.search API...
   * both? if MN search is not implemented, only give search CN option
   * give user choice which objects search
 * How do we deal with variable sync lag times on CNs?


09 Aug 2012 Additional discussion
-------------------------------------------------

 * Data Packaging in Morpho