Re: [dam-cmis-profile-discuss] RE: Key Objectives For CMIS/DAM

[David Choy <david.choy500@gmail.com>]

I wrote the following offline yesterday. It would complement Florian's post.

---

CMIS is a generic, interoperable, client-server interface which enables a client to manage content stored in a CMIS-compliant content management repository. A CMIS repository is logical. The CMIS interface may be mapped (typically by a system provider or a service provider) onto a physical content management system. Many popular content management systems have supported a CMIS interface on top of their native interface (e.g. Filenet, Documentum, Sharepoint, Open Text, Alfresco, and others). There are open source implementations of CMIS, such as the Apache Chemistry project which has implementation for both server and client.

The CMIS specification can be separated into two parts. The first part defines the abstraction for a CMIS repository. It includes a repository data model and a set of services that a client may use to access a CMIS repository. The second part describes the protocols a client actually uses to communicate with a CMIS repository. CMIS leverages popular protocols instead of inventing a new one. Since the popularity of communication protocols changes over time, the core of the CMIS specification is by design decoupled from any specific communication protocol. Currently, CMIS defines protocol bindings for SOAP-based Web Services (a service-oriented protocol), for AtomPub (a resource-oriented RESTful protocol), and for browsers (leveraging the capabilities that a modern web browser has). Each binding describes how CMIS information is passed/interpreted using the respective protocol. A CMIS client may use any of the three bindings to access a CMIS repository.

The CMIS data model is based on the concept of typed objects. An object may be a standalone information entity (such as a digital asset), a folder, a relationship between two objects, or a management policy. Each object has a system-assigned, unique (within the scope of the given repository), and persistent object id; has a fixed primary object type, and zero or more additive secondary object types; has a number of metadata properties; and optionally has content streams. Content streams are opaque to CMIS, and can be primary content, thumbnail, or other unstructured data. Much of the data model actually deals with metadata. A metadata property may be defined by CMIS, by a repository, or by an application or a user. A primary object type specifies a set of properties that each object of that type must have. In addition, an individual object, regardless of its primary object type, may have additional properties specified by the secondary type(s) that the object is associated with. During the life span of an object, secondary types may be added to or removed from the object. CMIS does not define NULL value for properties, but a property may be left unset. Some properties are maintained by system (such as an object's creation time). Others may be "required" (i.e. they cannot be left unset). An object type definition specifies all the attributes of an object type, plus a list of property definitions which in turn specifies all the attributes of each property. From object type definitions, an application can programmatically discover what kinds of objects there are in a repository. Furthermore, an application can create new object types to accommodate application-specific metadata.

CMIS supports type hierarchies with single inheritance. Each subtype of an object type inherits all the property definitions of its parent type, and may have additional property definitions of its own. Five root primary object types are specified by CMIS: Document, Folder, Relationship, Policy, and Item, each carrying a base set of properties for CMIS to function properly. An application may, for example, create a new object type called "Book" as a subtype of "Document", inheriting all its properties, plus additional properties: "Authors", "Publisher", etc. It may further create another object type "Tour Book" as a subtype of "Book", inheriting all its properties, plus additional properties such as "Geographic Region". However, creation of a new root primary object type is not allowed, but creation of new root secondary object types is allowed.

Besides standalone info objects (the "Document" type and its subtypes), CMIS supports a hierarchy of folder objects. An info object may be filed into one or more folders, or left unfiled. Furthermore, a binary relationship between two objects may be represented by a relationship object and described by the relationship object's properties. Multiple relationships may exist between two given objects. Policy objects are used to hold management policies which can be applied to other objects. Item objects may be used to represent entities that do not fit the mold of the other four kinds of objects.

CMIS services cover the basic CRUD operations (Create, Retrieve, Update, Delete), navigation (through the folder hierarchy and relationship objects), and SQL-based query with extensions to support fulltext search and scoping by folder. Other CMIS features include object retention and hold, ACL-based object-level access control, and document versioning.

David

On Fri, Nov 1, 2013 at 2:10 PM, Mueller, Florian <florian.mueller02@sap.com> wrote:

Let me try to explain the CMIS 1.1 specification in a few sentences.

The CMIS specification is divided into two parts. The first part describes
the domain model (chapter 2) and the second part the bindings (chapters 3,
4, and 5).

The domain model defines the six base types (document, folder,
relationship, policy, item, and secondary), their metadata and their
behavior (section 2.1). It also describes all CMIS operations (50+)
(section 2.2). The domain model does not say how a client and a server
communicate, though.

The CMIS bindings chapters define how the domain model is mapped to the
bytes on the wire. CMIS 1.1 defines three bindings:
- The Web Services binding, based on SOAP (chapter 3)
- The AtomPub binding, based on the Atom specification (chapter 4)
- The Browser binding, based on JSON and form data for _javascript_
applications in a web browsers (chapter 5)

The bindings are equivalent and a client can choose the binding that works
best in its environment. (There are only a few minor things you can't do
with the AtomPub binding.)
There are client libraries for many programming languages today that make
the bindings (almost) transparent to a developer. A developer uses the API
of the client library and eventually only needs to learn the domain model.

The domain model defines a set of mandatory data structures and operations
and several optional capabilities. A client can ask for the so-called
"repository info" to determine which capabilities the server provides.
The repository info has an extension point that would allow exposing
profile information. For example, a DAM server could provide information
about the DAM profile it supports and DAM specific capabilities. A DAM
client would be able to determine if the server supports the feature set
it needs.

A CMIS server is not allowed to expose another (seventh) base type, but it
can provide a hierarchy of types derived from the six CMIS base types.
These types can add more properties and may have a different behavior than
the parent type.
For example, a DAM profile could define that a server must provide a type
"dam:asset" that is derived from "cmis:document" (section 2.1.4). It would
inherit the properties of "cmis:document" and add new DAM specific
properties. The profile could also define that all objects that carry
content must be of the type "dam:asset" or a type derived from that. A
pure "cmis:document" object wouldn't exist in such a repository. The
repository would still be CMIS compliant and generic CMIS clients could
work with it.

I don't think there is anything to simplify if you start with the
mandatory CMIS capabilities and build the DAM features on top. Also the
Browser binding is a very simple and lean wire protocol. Find a few
examples here: [1]


I hope that helps,

Florian


[1] http://docs.oasis-open.org/cmis/CMIS/v1.1/os/examples/browser/

On 01.11.13 10:31, "Ralph Windsor" <ralph@daydream.co.uk> wrote:

>On point (1) of my message from last week, it seems to me that the
>cornerstone of any simplified variant of the CMIS protocol for DAM is a
>definition of the Objects to be stored (which are probably the assets in a
>typical DAM system in the vast majority of cases).  This should describe
>the
>core essential object properties which any compliant DAM systems need to
>be
>able to generate to allow more fully CMIS compliant tools to interpret
>them.
>
>I lack a full grasp of the CMIS protocol, is anyone on the list able to
>explain, in basic terms, how this works in CMIS and also cite any relevant
>sections from the CMIS spec?
>
>We'll almost certainly need to extend this with compatibility mappings
>between other protocols, but they can all come in later.  Right now, it's
>the real 'need to know' stuff we should  be interested in.
>
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail:
>dam-cmis-profile-discuss-unsubscribe@lists.oasis-open.org
>For additional commands, e-mail:
>dam-cmis-profile-discuss-help@lists.oasis-open.org
>


---------------------------------------------------------------------
To unsubscribe, e-mail: dam-cmis-profile-discuss-unsubscribe@lists.oasis-open.org
For additional commands, e-mail: dam-cmis-profile-discuss-help@lists.oasis-open.org