We've been talking a lot about how datastore version 3 (?) should be
structured.  I'd like to propose (purely to initiate discussion) that it
be structured as follows:

The datastore is a collection of objects, which are arranged into version
trees.  Each object has a tree_id, which is an arbitrary unique identifier
(string) for the version tree.  The datastore will generate a new tree_id
each time a new tree is created.  Each object also has a version_id, which
indicates the object's place in the tree.  The version_id could take the
form of a dotted decimal string like "4.5.2.1".

Some objects are Actions, and some objects are Documents.  Each Action
must retain a list to all Documents associated with that Action, each
represented as a (tree_id, version_id) pair.  Each Document must retain a
reference to the Action with which it is associated, represented as a
(tree_id, version_id) pair.  Each Document is only associated with a
single Action; a new Action that uses a Document must always make a new
version.

To implement the concepts present in the current datastore, activity_id
would actually be the tree_id of an Action, and object_id would be the
(tree_id, version_id) pair of a Document.

--Ben

Attachment: signature.asc
Description: OpenPGP digital signature

_______________________________________________
Sugar-devel mailing list
Sugar-devel@lists.sugarlabs.org
http://lists.sugarlabs.org/listinfo/sugar-devel

Reply via email to