Public review comments for xmile-v1.0-csprd01

On behalf of the developers of Simantics System Dynamics (SimanticsSD), I would like to provide the following comments for the public review of the specification draft (consolidated to one message to minimize unnecessary clutter in the mailing list, hopefully this is okay).

3.2.2.1 Identifier Form:

[technical / minor] Since underscores are treated as whitespace even inside quotation marks, is it possible to add an escape sequence for the underscore character? It seems somewhat odd that even though the identifier can contain almost any Unicode characters, literal underscores are not expressible with the current specification.

3.2.2.3 Namespaces:

[technical / minor] We understand that Simantics System Dynamics is (as of this writing) a relatively small player in the field of system dynamics modelling and simulation software vendors, but as the tool is open source and freely available and we are planning to implement support for XMILE as soon as the specification is finalized, would it be possible to have a “simanticssd” namespace added to the list of vendor namespaces?

3.7.1.1 Array Operations

[editorial or technical / major] The section mentions that XMILE operations on arrays should perform element-by-element operations instead of their linear-algebra analogs. The section could be expanded to also cover how the linear-algebra operations are expressed in XMILE, or, if the operations are not covered by the specification al all, could they possibly be added? In SimanticsSD we support both element-by-element operations (e.g. .*) and the linear-algebra counterparts (e.g. *) for array-valued variables, and even though the operations could technically be implemented as macros in the XMILE model, direct support would simplify things greatly.

4.4 Auxiliaries

[technical / minor] This is somewhat unrelated to the specification, but I ran into a strange issue when I tried to implement a prototype XMILE export in our tool; the Java Architecture for XML Binding (JAXB) binding compiler xjc (specifically version 2.2.4-2) fails to generate Java classes for the provided schema, because the element that represent auxiliaries is named “aux”. The tool fails to produce any errors and simply omits the generation of Aux.java, which results in an incomplete set of classes that does not compile. I have no idea why this happens, but renaming the element into “auxiliary” in the schema fixed the problem. I have not tried other versions of the tool yet but if the issue persists I shall have to file a bug report and hope it will eventually be fixed. Since JAXB is relatively widely used, I though you should be made aware of this issue.

4.7.1 Modules

[technical / major] According to the section, the module input and output definitions are defined as they are “to support arbitrary re-use of submodels”, but as the modules reference the submodels via either the name or the resource attribute and they do not have a separate name attribute of their own, it is not immediately clear what “re-use of submodels” means in this context. Is it possible to re-use submodels inside a single model, i.e. create several modules that reference the same submodel but are connected to different input and output variables in the parent model? This is again relevant in the context of our tool since our “modules” (sorry about the overloaded terminology) are user-defined elements with internal model structure (i.e. submodels) that can be freely instantiated in the parent model. This section should be expanded with a more defailed definition for “re-use”, and whether or not XMILE supports this sort of behaviour. (It should be noted that even though XMILE would only support one module for one submodel in a model, this would not prevent the implementation of XMILE support in SimanticsSD.)

Best Regards,

Janne Kauttio

xmile-comment message