[auscope-geosciml] Package dependencies in GeoSciML 3 UML model [SEC=UNCLASSIFIED]

Rob.Atkinson at csiro.au Rob.Atkinson at csiro.au
Mon Dec 6 19:50:19 EST 2010


A new connector has been created 'EarthMaterial -> CGI_Utilities' (A package link has been found between 'EarthMaterial' and 'CGI_Utilities')  [This "error" and the PhysicalProperties one are because the HollowWorld tool creates "Dependency imports" whereas we have until now been using "Package imports" in the UML.  Hence the HW tool does not recognise that the packages are already related.

Will look into this, chat with the gurus and decide what correct behaviour is. I think the tool should perhaps detect equivalent forms and likely errors and enforce a consistent idiom.

Can the HollowWorld tools for diagrams be made a bit smarter? eg, if a Context or Package Dependency diagram already exists, can the tool analyse what exists and add to it where necessary?  The tools are fantastic for building models from scratch, but can be a little frustrating when applied to a model like GeoSciML which is well developed.

The tool only adds where it didn't find a recognisable import - it's a matter of getting the forms of import right and updating similar forms to a consistent idiom I think.


 Also (for the benefit of others who were not privy to our conversation last week) due to  the fact that many ISO UML packages are not stereotyped as "Application Schema", the HW  tools (for Package diagrams and Context diagrams) appear not to be processing ISO App Schema dependencies properly.  Either the ISO UML or the HW tools should be updated.]


We are working to update the ISO models. It's impossible with an arbitrary level of nesting of packages to detect the intended "unit of reuse" with the current ISO models. I'm meeting with ISO HMMWG and Project Management Group today and tomorrow, and this will be on the agenda. SolidGround tools will recognise an isApplicationSchema tagged value, so hopefully this is a policy decision we can get through easily, otherwise they may want to make some more explicit mechanism. (IMHO they should use their own Conceptual Schema Language given its scope is supposedly ISO models!)

Rob
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opengeospatial.org/pipermail/geosciml/attachments/20101207/fa1802e1/attachment.htm>


More information about the GeoSciML mailing list