Simulation Data Model Documentation

[Rick Wagner]

Dear Laurent,

> First, I am happy that this mailing list is active with many  
> contributions
> recently.
> Besides, the simulation data model changed a lot thanks to Gerard and
> other contributors in a good way I think.
>
> Since last InterOp (end of september), the public documents about  
> SNAP DM
> & protocol did not change at all,
> so I tried to use them & implement the SNAP DM for french simulations
> (galmer, horizon halo & galaxy catalogs, PDR...) and this work is  
> still
> going on.
>
> I want now to insist on one main point : SNAP documentation &  
> change logs.
>
> Gerard changed SNAP pages to provide a new UML model, updated XML  
> schemas
> and a new online documentation appeared : great !
>
> However, does it imply that the previous reference document
> (SNAP_SimulationDM.doc) will not be updated anymore ?
>
> My concrete problem is to find differences between previous and  
> current
> versions of the SNAP DM.
>
> I purpose to use UML for what it is efficient = modelling.
>
> I mean the reference document for SNAP DM should be the UML model  
> (edited
> with the community edition of Magic Draw UML)
> and provided as a single xml file : SNAP Simulation DM.xml.
>
> I think that xml schemas & online documentation should be generated  
> from
> this xml document mainly by XSLT transformations
> or any other solution (maybe, ask help to DAL group) because it  
> would be
> more accurate & easy to maintain.
>
> For now, the problem is that the UML project does not contain  
> following
> information :
> - short & long description of every entity (SNAPProtocol,  
> SNAPExperiment ...)
> - cardinality of attributes & relations : optional or required  
> attributes
> & collections should be defined
> - UML model & xml schemas are not synchronized
> - xml schemas do not contain some classes (for example :
> SNAPPostProcessing hiearachy), but have detailled descriptions for
> entities & attributes
>
> So it is difficult to get descriptions and details of SNAP entities &
> attributes,
> because the information is spread out in UML document & xml schemas.
>
> Besides, it is hard to maintain manually those documents I think,
> that's why I purpose to use the UML model as the 'Reference'  
> document with
> as much information as possible.
>
> Finally, a change log is needed to provide details on DM evolutions as
> Gerard begun to do it.
> Maybe, it could be generated too by a difference analysis tool...


I think the start of a solution to what you're suggesting a version  
control system (Subversion gets my vote), and perhaps something like  
a Trac site. This would give easy access to a change log, and a way  
to compare versions. Both the NVO and AstroGrid use Trac and  
Subversion, so I suspect someone would host the project. If not, I  
will offer up space on our server.

Unfortunately, I am much less informed about UML, and the possible  
conversion from XMI (the XML format for UML) to XML Schema.

--Rick