state table clarifications

["Peter Furniss" <peter.furniss@erebor.co.uk>]

These are summaries of the state table peculiarities that I found in the analysis that look likely to remain. I have not raised them as issues, as the point is not whether I think they should be fixed, but whether the committee thinks leaving them alone will risk public review comment.  So if anyone wants to take them up, they are most welcome.

 

A) CV table, Comms Times Out/Aborting - would be better as Resend Rollback/Aborting.

 

As it is, if a message is lost the state machines get stuck, with the coordinator in Aborting, participant in None. The coordinator needs to be able to re-solicit a response from the Participant. (this relates to the welcome side-effect of resolving 081, CV Aborting is now exclusively the "Rollback sent, Aborted expected" state)

 

B)  PV table, "Commit Decision" has two different meanings. It ought to be two different events - leaving as it is risks comment from the perplexed, as the naive reader will assume one named event is always the same thing.

 

C) If the 090 resolution leaves PV All Forgotten row still there, the  All Forgotten/None cell should be N/A, since the (non-existent) state machine doesn't experience Internal Events.

 

B and C are clarifying editorials, with no impact on behaviour.  I believe A reflects intended behaviour which is currently not shown in the table.

 

Peter