MARTE — MARTE name prefixes

There is no consistent convention followed in the beta document for naming of stereotypes and their attributes, and this has been pointed out in other issues. Some chapters use a prefix for the chapter, such as Pa or pa for performance analysis, Grm or grm for General Resource Model, etc. MOst do not.

From the point of view of precision, fully qualified names are sufficient, but from the point of view of usability, the prefixes are very helpful, especially as MARTE is such a big profile and realtime is a cross-cutting concern. THat is (unlike sysml) the profile does not redefine UML or restrict it to the domain of realtime, but adds realtime info to a functional design. MARTE may be used with other profiles, possibly several at once.

The usability issues are to identify the realtime stereotypes and attributes, and also to avoid cluttering the namespace. A stereotype attribute may become a property of an object and its name might conflict with a functional property. Users might avoid MARTE just to avoid the risk of this happening.

Issue: Consistency in prefixes is desirable, either all stereotypes with prefix, or none. Usability is affected. Realtime concerns are pervasive and crosscutting in software, so flooding of the namespace could be a problem, including conflict with other profiles and with functional property names.

Proposed resolution:
Apply two or three prefixes to all stereotypes and attributes:

mm for MARTE Modeling
ma for MARTE Analysis
possibly mc for MARTE Core.

Where a stereotype is specialized from another one, a third letter could be used.

Alternative proposal:
Apply prefixes only to the analysis chapters, because these are more likely to be used outside the embedded system domain (e.g. for enterprise system QoS) and thus to be used with other stereotypes. For instance to specify requirements on delay.

In order to decide when we need to prefix the name of element, we propose to
apply the following algorithm as much as possible:

• When we define a stereotype which name may conflict with the UML2
  metaclass the stereotype is extending. In this case, we prefix with the
  name of the sub-profile wherein the stereotype is defined.
• When we define a stereotype generalizing an already defined stereotype
  of MARTE. In this case, we should use the name of the profile wherein the
  stereotype is defined to prefix this latter.
• Finally, we should not use "_" between the prefix and the name of the
  prefix. This rule should apply also for the name of the concepts of the
  domain model. However, there is an exception to this rule: we will not
  rename the predefined NFP type (the one define in the MARTE library
  such as NFP-Real) because they are used all along the specification and
  the change will need too much work w.r.t. to the global interest to do it.
  Let's note that, we consider that that as an editorial issue and we will not provide
  the complete revised text for the resolution because it would be too big. The
  editor of the specification (Sebastien Gerard) will do the required modifications
  directly in the specification when integrating all other resolutions voted within this
  FTF2