Do we need gaps?

[MattWindsor91]

The current metamodel distinguishes between actions and gaps, the latter of which modifies the action by allowing some sort of indefinite communication before the action. The main reasons the two are rigidly separate are that:

• gaps are inspired by intraMSGs in property sequence charts (Autili et al. 2007), with (arrow) actions effectively being arrowMSGs (further, strict gaps are like the PSC 'strict operator');
• gaps attach onto an action and modify its semantics (currently invasively, but #7 might change that);
• it's not yet clear how gaps will appear in Sirius, and they might end up being annotations on action representations.

This question tracks the possibility of removing the distinction and treating gaps as a separate form of action.

Unsolved questions:

• [ ] If we switch to encoding sequential composition using ; (#8), we'll need a way of making sure that the gap can still modify an action rather than being composed alongside it. Maybe the action would, recursively, take an action; this would nicely reflect how the semantics works.
• [ ] Does this give us any concrete improvements, aside from simplifying the metamodel a bit?
• [ ] How would this look in a diagram?
• [ ] How would we calculate the communication set of a gap when deciding how to expand the RUN of a gap that, itself, wraps a gap...?

[MattWindsor91]

Now that I'm fairly convinced gaps are closer to PSC constraints than strict operators, I'm kinda interested in changing gaps into an action that wraps another primitive action, but I'd like to understand how this relates to other formalisms before taking the plunge.

[MattWindsor91]

Closing this as I think I've hit a stable state with gaps now: they specify what happens in the n tocks before an action.