Closed stanpinteTheSignallingCompany closed 10 years ago
In the version of the model evaluated, §3.5 is partialy implemented :
concerning true table, condition with high priority are not implemented (at least conditions 1, 5, 6, 10, 50) :
however, this is not a problem due to that other models of the benchmark cover only in part the high priority items, but it does not allow me to assess these criteria.
It is true that the Section 3.5 is not fully implemented, but it has enough implementation to show that timeouts are supported (see Stan's screenshot).
The priorities between the conditions of the Table 4.6.2 are fully implemented by the procedure Kernel.ModeTransitions.UpdateMode. The conditions themselves are implemented by the corresponding functions of the namespace Kernel.ModeTransitions. Only the conditions in blue on your screenshot are not implemented, so the conditions 1, 5, 6, 10 and 50 are implemented.
There is the following note:
"The current model does not cover the § 3.5 on session management and the §4, thus it is di cult to evaluate the capabilities of the language to treat time-outs and truth tables."
This statement is incorrect, as a good deal of §3.5 is implemented, including timeouts (see for instance this screenshot for timeouts).
Also, the high-priority elements that should be modelled to prove the support of truth tables (HIGH PRIORITY §4.6.2 (Transitions Table) and §4.6.3 (Transitions Condition table)) have been 100% and successfully modelled.
Therefore, could you please: