Here it is included what is in our intention the final version of the document. There is only one change from the previous version. The attribute Tags has become the extension Tags, and has been moved to the corresponding section. The reason for this is that adding an atribute cause backwards compatibility problems with some existing APIs to access the AC and so, in the interest of backwards compatibility, the data has been moved to an extension. Bye, Vincenzo Vincenzo Ciaschini wrote:
Blair Dillaway wrote:
Vincenzo,
In reading this draft I've found a few places I believe would benefit from some additional clarification.
(1) 3.1.1 - "As a consequence of this, in VOMS ACs the only admissible choice for the field is the baseCertificateID."
You might re-write this to say the holder field MUST include the baseCertificateID field and omit the entityName and objectDigestInfo fields. The syntax defines a sequence of 3 optional fields, not a choice.
But the RFC recommends that only one should be used. Done.
(2) 3.4.1.2 - "Where <root group> is by convention the name of the virtual organization."
The document seems to imply this is a required encoding for interoperability purposes. If so, its not just a convention. You should clarify if this is a MUST, SHOULD, or RECOMMENDED encoding.
Right, done.
(3) 3.4.1.2 - "Future compatibility issue: It is possible that in the future a /Role=NULL component may be omitted in its entirety. The same goes for a /Capability=NULL part. Conforming applications SHOULD be prepared to handle these cases."
The previous paragraph states "The /Capability=<capability name> part is deprecated....". If so, I assume conforming implementations SHOULD always omit the /Capability part whether or not its null. Having it be deprecated and noting it may disappear in the future seem to be in conflict.
Unfortunately, that is the besr that ca be done for the moment, since it should continue to be included until the software out there has has a reasonable chance to change its implementation not to rely on it. Obviously, this come free if the official APIs are used.
Also the statement that /Role=NULL *may* be omitted in the future seems to be in conflict with the examples in 3.4.1.3. The compact format shown does omit Role=NULL. If this is allowed, then the 'in the future' qualification should be removed. It should be stated if the compact format is recommended, required, or simply a supported encoding option.
The intent of the sentence is to state that applications interested in reading the date should deal with the case that the field could be empty.
(4) 3.4.2.2 - "A name-specific syntax that encodes multiple values in a single pair is also allowed."
Examples of a single value and multiple value encoding would be helpful here. The syntax in 3.4.2.1 indicates a Tag includes only a single name, value, and qualifier field. I assume how multiple values are encoded into the value field should be specified for interop purposes.
As said, it is name-specific, meaning that each attribute can choose its own way.
Regards, Blair
Bye, Vincenzo
-----Original Message----- From: owner-ogsa-authz@ggf.org [mailto:owner-ogsa-authz@ggf.org] On Behalf Of Vincenzo Ciaschini Sent: Friday, April 28, 2006 6:34 AM To: ogsa Authz Subject: [OGSA-AUTHZ] Updated version of VOMS Specification
Hi to All,
This is the updated document about VOMS specification I promised yesterday.
Apart from a few minor language clarifications, the main changes are in the following sections:
3.4.1 clarifications about the syntax of FQANs. 3.4.2 rewritten, with a (slight) change in format. 3.5.1 explanation rewritten. The previous explanation was the exact opposite of the truth. OOPS!
Also, the AC example in section5 has been substituted with a more complete one.
Bye, Vincenzo