[occi-wg] Core errata

[Ralf Nyren]

On Tue, 25 Sep 2012 08:04:39 +0000, "Feldhaus, Florian"
<florian.feldhaus at gwdg.de> wrote:
> Dear all,
> 
> Andy and I had a discussion on IDs for OCCI Entities last week wether
they
> should be UUIDs and wether they have to be globally unique. In the
> published version of OCCI Core, it states that entity IDs must be URIs
and
> "A unique identifier (within the service provider’s name- space) of the
> Entity sub-type instance.". In 3.1 of the OCCI RESTful HTTP Rendering
> document, it says:
> "Each resource instance within an OCCI system MUST have a unique
> identifier stored in the occi.core.id attribute of the Entity type [1].
It
> is RECOMMENDED to use a Uniform Resource Name (URN) as the identifier
> stored in occi.core.id.
> 
> The structure of these identifiers is opaque and the system should not
> assume a static, pre-determined scheme for their structure. For example
> occi.core.id could be urn:uuid:de7335a7-07e0-4487-9cbd-ed51be7f2ce4."
> 
> In my opinion it would be very good if both sentences could be moved to
> OCCI Core to clarify the ID format. Thus we would stress the fact, that
> UUIDs should be used, but we probably shouldn't require that they have
to
> be used. It is probably not necessary to enforce IDs to be globally
unique,
> but with UUIDs implementers can follow the guidelines in the RFC to
ensure
> global uniqueness.

+1

For the sake of backward compatibility I agree we probably cannot require
Entity.ID to be a UUID. A "SHOULD" is an acceptable compromise IMO.

Just to clarify, do you think that OCCI Core should recommend a specific
format of the Entity IDs as well?

I mean, even if Entity.ID is a UUID there are multiple ways to represent a
UUID. E.g. canonical form, binary format, URN, etc.

Having a specific ID format specified directly in OCCI Core would
definitely help with consistency across renderings. However from a
technical perspective I think the particular UUID format to use should be
up to the rendering.

regards, Ralf

> Am 03.09.2012 um 19:01 schrieb Ralf Nyren:
> 
>> Dear all,
>> 
>> I have update the Core document to include the changes discussed on the
>> last few confcalls.
>> 
>> The LaTeX changes are available from the core-errata branch in the
>> occi-wg
>> Git repository. Please find a compiled document attached.
>> 
>> 
>> Add a Attribute type to the Core Model
>> --------------------------------------
>> 
>> Many who have implemented the OCCI Core model already added an
Attribute
>> type. This change formalise the notion of "attribute properties".
>> 
>> Instead of just saying an attribute have a name, an attribute now has
its
>> on Attribute type with both name an properties.
>> 
>> The new Attribute type is an *identifier* for resource attributes. It
>> does
>> not contain the attribute value. I.e. just as before.
>> 
>> A distinction has been added between "client discoverable attributes"
>> (called OCCI Attributes) and attributes internal to the model (called
>> model
>> attributes). occi.core.id, occi.core.source and occi.core.target have
>> been
>> changed back to the model attributes they were until shortly before
HTTP
>> Rendering was released. These are determined to be model attributes and
>> thus receive special care in renderings.
>> 
>> 
>> Change Action to inherit Category
>> ---------------------------------
>> 
>> The concept of Action representing the "invocable operation" itself is
>> removed. Instead an Action is just an *identifier* of the operation.
>> 
>> An Action instance identifies an invocable operation in much the same
>> way as a Kind instance identifies an Entity sub-type.
>> 
>> Impact on the existing HTTP Rendering (occi/1.1) is none. In fact the
>> text/ renderings already use "type=action" in its Category headers.
>> 
>> regards, Ralf<core.pdf>_______________________________________________
>> occi-wg mailing list
>> occi-wg at ogf.org
>> https://www.ogf.org/mailman/listinfo/occi-wg