<html>
  <head>
    <meta content="text/html; charset=utf-8" http-equiv="Content-Type">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <div class="moz-cite-prefix"><br>
      I just bring my 2c<br>
      I think we have to keep in mind that the goal of having two
      independent implementations for validating a new IVOA proposal  is
      not only to demontrate that the idea is well defined and coherent,
      but also to check that this idea fulfill a real need. Finding two
      independent institutes implementing a protocol, ideally as a
      provider &amp; consumer, has always been an excellent method to
      check this real need. Otherwise, you will just have one - from the
      author of the idea.<br>
      So, do we have to validate this real need in the case of DM ? and
      if yes, how to do it ? So I would share the Gerard idea that the
      validation of the DM should be based on the implementation of the
      use cases for which de DM has been defined. And ideally by two
      independent institutes as provider &amp; consumer.<br>
      <br>
      Cheers<br>
      Pierre Fernique<br>
      <br>
      <br>
      Le 10/05/2016 18:04, Gerard Lemson a écrit :<br>
    </div>
    <blockquote
cite="mid:CAPoLd+zRK_O_VmGApGbJGHi8F-xYDZXuK53AyrTGKBPz3aJxuw@mail.gmail.com"
      type="cite">
      <div dir="ltr">HI Tom
        <div>I think that writing a model in VO-DML is *not* an
          implementation of the model, but its definition. It *is* an
          implementation of VO-DML,</div>
        <div><br>
        </div>
        <div>An *instance* of the model is a valid implementation of the
          model, but we have no generic standard way (yet) for
          representing instances of data models. The mapping document
          will fit that, but protocols can also do that.</div>
        <div>Showing interoperability of the data model could be an
          application that uses two or more independent instances of the
          data model serialized in some standard way.</div>
        <div>One way this might happen is that votables annotated with
          the same data model are interpreted as instances of the data
          model.</div>
        <div>And it would be nice if something interesting is done with
          them. Ieally this could be an implementation of a use cases
          the model was supposed to support.</div>
        <div><br>
        </div>
        <div>Gerard</div>
        <div class="gmail_extra"><br>
        </div>
        <div class="gmail_extra"><br>
          <div class="gmail_quote">On Tue, May 10, 2016 at 10:46 AM, Tom
            McGlynn (NASA/GSFC Code 660.1) <span dir="ltr">&lt;<a
                moz-do-not-send="true"
                href="mailto:tom.mcglynn@nasa.gov" target="_blank"><a class="moz-txt-link-abbreviated" href="mailto:tom.mcglynn@nasa.gov">tom.mcglynn@nasa.gov</a></a>&gt;</span>
            wrote:<br>
            <blockquote class="gmail_quote" style="margin:0 0 0
              .8ex;border-left:1px #ccc solid;padding-left:1ex">If you
              feel that data models are not subject to the requirement
              of two reference implementations, that's fine but then
              this discussion is moot regardless.  If you think they are
              then you are quibbling about my choice of words. Feel free
              to substitute whichever you like for 'protocol'.<br>
              <br>
                  Regards,<br>
                  Tom<br>
              <br>
              Matthew Graham wrote:<br>
              <blockquote class="gmail_quote" style="margin:0 0 0
                .8ex;border-left:1px #ccc solid;padding-left:1ex">
                But a data model is not a protocol.<br>
                <br>
                -- Matthew<br>
                <br>
                On May 10, 2016, at 4:37 PM, Tom McGlynn (NASA/GSFC Code
                660.1) wrote:<br>
                <br>
                <blockquote class="gmail_quote" style="margin:0 0 0
                  .8ex;border-left:1px #ccc solid;padding-left:1ex">
                  This kind of using the fact that you have written a
                  definition of the model counting as an implementation
                  of the model sounds awfully incestuous. I have always
                  read the requirement as having two different groups
                  using the protocol in some service, ideally one that
                  supports doing astronomy.<br>
                  <br>
                      Tom McGlynn<br>
                  <br>
                  Matthew Graham wrote:<br>
                  <blockquote class="gmail_quote" style="margin:0 0 0
                    .8ex;border-left:1px #ccc solid;padding-left:1ex">
                    So once we have a mapping standard with reference
                    implementations any DM model specified in VO-DML
                    could automatically have a reference implementation
                    (according to these criteria) which would make life
                    easier. Time to get that mapping spec out :)<br>
                    <br>
                    -- Matthew<br>
                    <br>
                    <br>
                    <br>
                    On May 10, 2016, at 4:00 PM, Laurino, Omar wrote:<br>
                    <br>
                    <blockquote class="gmail_quote" style="margin:0 0 0
                      .8ex;border-left:1px #ccc solid;padding-left:1ex">
                      Hi,<br>
                      <br>
                      I agree with Gerard that items 1) and 2) look like
                      the same, unless one brings in a standard for
                      instance serializations. Since we can have
                      standardized mapping strategies (as different
                      recommendations) that map instances of valid
                      VODML/XML models and their standard
                      serializations, I don't think a valid
                      serialization of an instance should be required
                      for data models: this should be guaranteed by the
                      mapping standard(s) and their reference
                      implementations.<br>
                      <br>
                      As for the ModelImport requirement, it makes sense
                      for "low level" models but not for "high level"
                      ones, plus there is no way to guarantee that all
                      types defined by a model are extendable/usable by
                      other models. It gets too complicated.<br>
                      <br>
                      I would suggest we include the "model import"
                      evidence as a "soft requirement", to be evaluated
                      on a case-by-case basis depending on the use cases
                      of the model. For STC, it makes a lot of sense to
                      require this additional proof of interoperability,
                      because the model is intended to be a building
                      block for other models.<br>
                      <br>
                      Or, we might be to *require* that at least one
                      reference implementation is a
                      mission/archive/service-specific model that
                      extends the standard one. So, if we had a model
                      for Sources/Catalogs, at least one reference
                      implementation should be a
                      mission/archive/system-specific model that proofs
                      the model can be meaningfully extended by actual
                      specializations. Other than properly validating as
                      VODML/XML, this model should be evaluated for its
                      domain-specific content, and that's probably not
                      something you can automate. This should also be a
                      good way to involve implementors from the
                      community, so it's probably my preferred one.<br>
                      <br>
                      Omar.<br>
                      <br>
                      <br>
                      On Tue, May 10, 2016 at 9:38 AM, Gerard Lemson
                      &lt;<a moz-do-not-send="true"
                        href="mailto:gerard.lemson@gmail.com"
                        target="_blank">gerard.lemson@gmail.com</a>
                      &lt;mailto:<a moz-do-not-send="true"
                        href="mailto:gerard.lemson@gmail.com"
                        target="_blank">gerard.lemson@gmail.com</a>&gt;&gt;
                      wrote:<br>
                      <br>
                          Hi<br>
                          What is meant by item 2, "An XML serialization
                      of the DM".<br>
                          The standard representation (serialization?)
                      of a VO-DML data<br>
                          model is VO-DML/XML, i.e. XML. And that is the
                      representation<br>
                          that can be validated (step 1?) using
                      automated means, for<br>
                          example using XSLT scripts in the vo-dml/xslt
                      folder on volute@gavo.<br>
                          If 2) is meant to imply an XML serialization
                      of an *instance* of<br>
                          the model, that we can only do once we have a
                      standard XML<br>
                          representation of instances of models. That
                      does not yet exist.<br>
                          The original VO-URP framework does contain an
                      automated XML<br>
                          Schema generator for its version of VO-DML,
                      that has not yet been<br>
                          ported to VO-DML.<br>
                          And of course the mapping document describes
                      how one can describe<br>
                          instances serialized in VOTable, but that is a
                      different standard.<br>
                      <br>
                          For what it's worthy, I think that an
                      "implementation of VO-DML"<br>
                          is a data model expressed using that language
                      (in VO-DML/XML to<br>
                          be precise) and validated using software. The
                      latter enforces<br>
                          that the language should allow automated
                      validtion.<br>
                          I think interoperable implementations of
                      VO-DML are two or more<br>
                          valid models that are linked by "modelimport"
                      relationships.<br>
                          I.e.one model "imports" the other(s) and uses
                      types from the<br>
                          other as roles or super types in the
                      definition of its own types.<br>
                          This is supported by the VODMLID/VODMLREF
                      meachanism of the language.<br>
                      <br>
                          Cheers<br>
                          Gerard<br>
                      <br>
                      <br>
                          On Tue, May 10, 2016 at 9:04 AM, Matthew
                      Graham<br>
                          &lt;<a moz-do-not-send="true"
                        href="mailto:mjg@cd3.caltech.edu"
                        target="_blank">mjg@cd3.caltech.edu</a>
                      &lt;mailto:<a moz-do-not-send="true"
                        href="mailto:mjg@cd3.caltech.edu"
                        target="_blank">mjg@cd3.caltech.edu</a>&gt;&gt;
                      wrote:<br>
                      <br>
                              Hi,<br>
                      <br>
                              We're trying to define specifically what
                      would satisfy the<br>
                              reference implementation requirement for
                      an IVOA Spec in the<br>
                              context of a data model. The proposal is
                      that:<br>
                      <br>
                              (1) If the DM has been described using
                      VO-DML it can be<br>
                              validated as valid VO-DML<br>
                      <br>
                              (2) An XML serialization of the DM can be
                      validated<br>
                      <br>
                              so therefore is the combination of the two
                      sufficient to<br>
                              demonstrate the validity and potential
                      interoperability of<br>
                              the data model (which is the purpose of
                      the reference<br>
                              implementations).<br>
                      <br>
                                      Cheers,<br>
                      <br>
                                      Matthew<br>
                      <br>
                      <br>
                      <br>
                      <br>
                      <span class="HOEnZb"><font color="#888888">
                          <br>
                          -- <br>
                          Omar Laurino<br>
                          Smithsonian Astrophysical Observatory<br>
                          Harvard-Smithsonian Center for Astrophysics<br>
                          100 Acorn Park Dr. R-377 MS-81<br>
                          02140 Cambridge, MA<br>
                          <a moz-do-not-send="true"
                            href="tel:%28617%29%20495-7227"
                            value="+16174957227" target="_blank">(617)
                            495-7227</a><br>
                        </font></span></blockquote>
                  </blockquote>
                </blockquote>
              </blockquote>
              <br>
            </blockquote>
          </div>
          <br>
        </div>
      </div>
    </blockquote>
    <br>
  </body>
</html>