Thanks for clarification! Good to know that it already has been fixed in the spec.
-----Original Message----- From: Florian Müller [mailto:[email protected]] Sent: Mittwoch, 12. Oktober 2011 23:35 To: [email protected] Cc: Jens Hübel Subject: Re: Not set properties Hi Jens, It has been clarified in the CMIS 1.0 errata [1]. If no property filter is set, the repository decides which properties to return. It could decide not send version properties for non-versionable documents. If the property filter is set ("*" or property list), then the client is responsible for all bandwidth waste. - Florian [1] http://docs.oasis-open.org/cmis/CMIS/v1.0/errata-01/csprd01/cmis-spec-v1.0-errata-01-csprd01.html#_Toc295214661 ----- Original Message ----- From: "Jens Hübel" <[email protected]> To: [email protected] Sent: Wednesday, October 12, 2011 9:33:12 PM GMT +00:00 GMT Britain, Ireland, Portugal Subject: RE: Not set properties Hi Chemistries, I am just reading through this older thread here. > -----Original Message----- > From: Florian Müller [mailto:[email protected]] > Sent: Montag, 22. August 2011 15:58 > To: [email protected]; Weigel, Achim > Subject: Re: Not set properties > > Hi Achim, > > Not-set properties have to be returned (see CMIS spec). There must be a > PropertyData object for each property. The value can be either null or an > empty list. OpenCMIS understands both. Can someone point me to the relevant section in the spec that not set properties have to be returned? Section 2.1.2.1 If a value is not provided for a property, the property is in a "value not set" state. There is no "null" value for a property. Through protocol binding, a property is either not set, or is set to a particular value or a list of values. 4.2 Web Services Binding Mapping Optional parameters and optional return values are not set if they are missing or their value is NULL. Do I read the first section correctly that this is binding specific? Do I read the second section correctly that this is optional for web services? I feel I missed finding the relevant phrase, but where is it? And does someone recall the motivation behind that? There are use case where you may have dozens of properties on a type where only few of them are actually used. This is a huge waste of network bandwidth then. Why do I have to return all the irrelevant versioning properties if my repository implementation does not even support versioning? Does this make sense? And should we change something or make this more explicit in the CMIS spec for 1.1? Jens P.S. At some point we should think about creating a FAQ...
