Ed, Thats a very good point. We need some form of higher level view of what the different implementations offer in terms of spec compliance and tests to demonstrate this compliance. Interoperability against this stated behaviour is, as you say, the sensible objective. Have any kind of functional matices been extracted from the specs by the project or spec team so that we can document compliance levels for each implementation? If not we could make one and try and get a current picture of where we are against the specs as they stand. This would also be useful as I expect the specs will continue to evolve and tracking progress against these changes could be difficult. Having this resource then makes it much easier to do a propoer job of the interoperability testing.
Regards Simon On 5/3/06, Edward Slattery <[EMAIL PROTECTED]> wrote:
On the subject of interop, what are we going to do about some form of compliance suite? We will need to define the range of interop expected, because I guess both Java and C++ implementations interpret things which are not covered by the spec. I am quite sure we do this in different ways. I think it would only be sensible to test interop within the specified behaviour, and highlight as we go along where the specification is open to interpretation. cheers, Ed. On 03/05/06, Edward Slattery <[EMAIL PROTECTED]> wrote: > > Can the Java SDO handle the Primer sample without moidification? The C++ > one certainly cannot as it uses ecore: stuff , which I dont have. > > cheers, > Ed. > > > On 03/05/06, kelvin goodson <[EMAIL PROTECTED]> wrote: > > > > There's a pair of files called "Primer.xsd" and "PrimerSample.xml" which > > I > > have used in the past which come from Eclipse EMF and > > is as comprehensive in its coverage of schema as I have ever > > needed. I'm > > not sure how we stand on being able to distribute these as part of the > > test > > suite in Tuscany. A quick google shows the files available from the > > article > > > > > > http://www.eclipse.org/newsportal/article.php?id=3666&group=eclipse.tools.emf > > as attachments > > > > http://www.eclipse.org/newsportal/attachment.php?group=eclipse.tools.emf&id=3666&attachment=2 > > and > > > > http://www.eclipse.org/newsportal/attachment.php?group=eclipse.tools.emf&id=3666&attachment=3 > > > > With regards to sharing change histories, I imagine the primary use > > case > > for change histories is when you give a give a modified graph back to > > the > > "same" DAS for writing back to the original source. So I in terms of > > cross > > language interoperability I would extrapolate that the scenario we would > > be > > supporting would be that of fairly tightly coupled DAS implementations, > > all > > accessing the same source. I may be wrong, but It doesn't sound like a > > frequently encountered scenario, so whilst it sounds like goodness, it > > wouldn't be at the top of my priority list. > > > > I don't have a strong feeling yet for the shape of tests we should do, > > but > > perhaps we could talk about which interop tests we want to do on the IRC > > channel in the next regular slot (I'm assuming that's Monday May 8th, > > at: > > 15:30 GMT, 16:30 BST, 08:30am PST, 11:30am EDT, 21:00 Bangalore) > > > > Cheers, kelvin. > > > > On 5/2/06, Simon Laws <[EMAIL PROTECTED]> wrote: > > > > > > I spent a little time getting familiar with PHP/SDO and had some more > > > thoughts about cross language interop testing. For SDO the most basic > > test > > > is to read and write an XML file from PHP/JAVA/C++ and compare the > > > results. > > > We could develop this to test other functions such as creating, > > updating > > > and > > > deleting content and again comparing the content from the different > > > implementations. Assuming we start with the same input we would expect > > the > > > outputs to be compatible. > > > > > > Is there an XML file somewhere that contains the full set of supported > > > > > types > > > and type constructs? > > > > > > Where relational DASs are implemented Similar tests could be carried > > out > > > with relational data ensuring that type conversions are performed > > > accurately > > > and consistently across implementations by reading out of a database > > and > > > printing out the data for comparison or by inserting back into a > > database > > > for comparison. > > > > > > We could also look at how consistently the implementations convert > > from > > > one > > > DAS type to another but I guess this is not strictly an > > interoperability > > > issue. > > > > > > Is there an intention that SDO change histories will be shared? If so > > this > > > is something else that you could expect to be transferred across > > language > > > boundaries and hence we should think about how to test this. > > > > > > Once the work is done to have C++ support Axis2 we should do some SCA > > > interop testing also. In the near term we could do some basic testing > > > based > > > on SDO/axiom conversions if it's thought that to be worth it. > > > > > > I'm happy to spend time setting up tests if we can agree which ones > > are > > > required. > > > > > > Thoughts > > > > > > Regards > > > > > > Simon > > > > > > > > > > > > -- > > Best Regards > > Kelvin Goodson > > > > >
