Re: oldssn:Device in new SSN and SOSA

Dear Pawel,

Thanks for your interest in the SSN new and thank you for your commendation.

Please do let us know of an early implementation of the use of SSN new as we are currently collecting implementation evidence (until the 8th of August).

I have forwarded your email to the public mailing list as it raises questions that others surely have too.

To answer your questions. There were several reasons for the removal of the Device and the SensingDevice class and we had several discussions on that issue. Please have a look on our wiki for more details and issues that have been raised in that regard. The following two pages are an entry point to that discussion:

https://www.w3.org/2015/spatial/wiki/Platform

https://www.w3.org/2015/spatial/wiki/Link_between_platform_and_device

while the following two closed issues point to long discussions on that subject:

https://www.w3.org/2015/spatial/track/issues/88 
https://www.w3.org/2015/spatial/track/issues/153 

Summarising, the Platform class in SOSA now largely plays the role of the old Device class, as a Platform is now not necessarily being physical. In the old SSN, a Platform was a Physical Object, whereas now it is just a DUL Object. As such, there is no need any more for a Device class, in particular, since SSN new would need several device classes since an Actuator Device would have different properties to a Sensor Device. As you rightly pointed out, we did remove the “wrongly” named SensingDevice class for that reason. We also removed the Device class for simplicity as it just extended the System class with no discernible difference.

In terms of the mapping to old SSN. We could not define ssn:System equivalent to oldssn:System, as the old System was a PhysicalObject, whereas the new System can also be a human, software or simulation (since sosa:Sensor is a subclass of System) and therefore maps to a dul:Object too. Sosa:Sensor is not equivalent to the oldssn:Sensor for the same reason. What that means for you is that you would still use the Sensor class as you have been using it previously. If you had used the Device class you can either use the System class (if you want the stronger SSN semantics) or use the Platform class (for the SOSA lightweight semantics). Example 8 shows the use of the Platform class in the context of an iPhone http://w3c.github.io/sdw/ssn/#iphone_barometer-sosa which could be a typical IoT use case.

In terms of naming, it would have probably been better to maintain the Device class compared to the System class of the Old SSN. However, the usage of the old Device class is narrower than the intended use of the System class, particularly regarding the introduction of an Actuator and a Sampler. Both these classes could not have been made a subclass of Device and therefore we opted for keeping the System class. Mapping the System class to IoT vocabulary would probably result in the label “Device”, though.

Hope that answers your question. One of the other members of the group may jump in if I missed something. I am sure we will address these issues in a paper to be written once SSN becomes a recommendation.

Kind regards,
Armin

On 18/7/17, 5:56 pm, "Pawel Szmeja" <pawel.szmeja@ibspan.waw.pl> wrote:

    Dear SSN editors and contributors,
    
    This is Paweł Szmeja from Polish Academy of Sciences, currently working 
    in the INTER-IoT project (part of EU IoT-EPI). First of all 
    congratulations on well-made update to the SSN ontology. We plan to make 
    use of it in our project and I have some questions and comments and I 
    hope you can help me clear things up, when it comes to my understanding 
    of SSN/SOSA.
    
    I was wondering why the Device class (now oldssn:Device) was removed 
    from SSN/SOSA. This class seems very useful when it comes to aligning 
    with other ontologies e.g. SAREF or even OneM2M Base Ontology. Moreover, 
    the suggested alignment of old SSN with SOSA states that an 
    oldssn:Device is a subclass of sosa:Platform and ssn:System. 
    oldssn:Sensor is, supposedly, equivalent to sosa:Sensor, which itself is 
    a subclass of ssn:System, but not necessarily of sosa:Platform. Doesn't 
    that lead to an "incompleteness" in class definitions? The suggested 
    alignment of oldssn, if taken as a whole, necessitates that sosa:Sensor 
    is a subclass of BOTH ssn:System and sosa:Platform, but the definition 
    of sosa:Sensor does not include subsuming sosa:Platform.
    
    If we take the sosa definition as more important, then not every 
    sosa:Sensor needs to be an oldssn:Device. To be frank I don't understand 
    why that would be the case in an ontology, generally speaking, dedicated 
    for IoT. This is especially puzzling to me, since the SensingDevice was 
    (rightfully) removed from the updated SSN.
    
    Can you please explain, or point me in the direction of an article or a 
    document that would explain the rationale behind removal of Device 
    class, and the fact that not every Sensor is a Device?
    
    Looking forward to your response.
    
    All the best,
    Paweł
    
    -- 
    Paweł Szmeja,
    Researcher,
    phone +48223810273
    Systems Research Institute Polish Academy of Sciences
    Newelska 6 01-447 Warsaw Poland

Received on Wednesday, 19 July 2017 02:40:01 UTC