Changes between Version 27 and Version 28 of GEC10InstMeasWorkingSession

03/13/11 10:21:33 (8 years ago)



  • GEC10InstMeasWorkingSession

    v27 v28  
    125125=== 4)  Measurement Data Objects and Descriptors (i.e, metadata) ===
    1261263:00pm [[BR]]
    127 Harry Mussman[[BR]]
    128   Read the following sections of the [  Architecture document: [[BR]]
     127Harry Mussman [[BR]]
     128  Read the following sections of the [  Architecture document:] [[BR]]
    130130  8.1    MEASUREMENT DATA OBJECTS (MDOS)         [[BR]]
    201201  Comments from earlier discussion: [[BR]]
     203  Camilo Viecco:[[BR]]
     205  All parameters look fine.  I have one question:  The 'Time' optional parameter appears to be redundant. Should it not match the owner/creator when mandatory value?[[BR]]
     207  There is also the issue of metadata for time ranges. How do we express that?  If we remove the time, then I belive this is a good place to start[[BR]]
     209  Hussam Nasir:[[BR]]
     211  I agree with Camillo suggestions and dont see any more addition/removal i would want to make.[[BR]]
     213  Max Ott:  [[BR]]
     215  As for meta data I first of all want to stress that we really should find a way to align this with the Resource Description spec of the Control Framework. While there is substantial reluctance there to embrace anything assembling some kind of structure, I'm not giving up the fight that easily. I just find it utterly silly to have two or more different ways to describe resources. Measurements are resources as well and metadata is all about describing where the measurements are coming from, that includes the resources we provision through the control framework.[[BR]]
     217  I know , with the little (or zero as in my case) funding we have available, nobody is excited about re-factoring for alignment sakes, but we shouldn't forget why we are we doing this in the first place. We just published a paper which included an analysis of last year's edition of one of the major conferences in our field. 26 out of the 36 accepted papers had substantial flaws in their respective analysis section, ranging from not even describing the experiment  to the sloppy "10-run average" without any indication of precision.  The numbers would have looked a lot grimmer if we would have tried to figure out on how to repeat the results.[[BR]]
     219  Coming back to meta data and repeating my RSpec argument. We first need an object/component/resource model and then we need a way to attach meaning to various 'groups' of model instances. The first part (model) should be quite stable as it gets baked into tools, services, and code and is expensive to change. The second part (semantics) should be easy to change as it inadvertently will as it reflects our changing needs, understandings, use cases. In other words, the rate of change somewhat reflects the vibrancy of the community using it.[[BR]]
     221  This is sitting in my drafts folder already for a few days, still half baked, but hopefully still of some marginal value.[[BR]]