posted 11 years ago
Hi Nikesh,
Each object in CMIS has a unique, immutable, and opaque object id. You can address every folder, document, document version, relationship, policy, and item (CMIS 1.1) with its object id. Object paths, on the other hand, are volatile and some objects cannot be reached by path.
The CMIS specification defines the following rules for objects.
Folders must have exactly one parent (except for the root folder) and have exactly one path.Documents, Policies, and Items can be filed (= one parent folder and one path), unfiled (= no parent folder and no path), or multi-filed (= multiple parents and multiple path).Older document versions may be filed, but that is not supported many repositories. Usually, document versions are not filed and don't have a path.Relationships cannot be filed and therefore never have a path.
This definition is not random. It reflects, more or less, the implementation of most content management systems. Operations that are based on the path of an object are often more expensive for content management systems than those that use the object id.
If you want to identify the same (or a similar) object in two different repositories, then neither the object id nor the object path helps you. The object id depends on the content management system and the instance. The folder hierarchy within a content repository also depends on the content management system and the administrator. In reality, it is virtually impossible to have the same path to a document in, lets say, Alfresco and SharePoint.
The key here is metadata. Define in both repositories a document type with properties. Lets say, an invoice has an invoice number. Then, you can perform a CMIS query like this:
Each repository would respond with the name and the object id the invoice document. The parent folders of documents don't matter. With the object id you can then perform whatever operation you need.
Regards,
Florian