Re: [ws-tx] Issue 102 - WS-AT: Editorial comments

[Ian Robinson <ian_robinson@uk.ibm.com>]

Monica.Martin@Sun.COM wrote on 31/10/2006 02:00:21:

>
> >Ian Robinson wrote: We never finished discussing this issue on the
> telecon so here are some email comments:
> >
> >First of all, we *did* start but not conclude the discussion on:
> >==>       Line 24: Change “Tentative actions visible to” to
> “tentative actions persistent and visible to”.
> >
> >
> >I agree that this simply observes an inconsistency in the text but, if
> >we're going to change it, I wonder if we should be generally more
precise
> >in this section. The AT Coordinator has no control over the persistence
or
> >the visibility of the Actions it coordinates through the registered
> >Participants - the AT Coordinator *only* provides the application with
> >atomicity. Making coordinated actions persistent and visible is the job
of
> >the Participants or the resource managers they delegate to. So I would
like
> >to suggest that we change the text at lines 19-26 from:
> >
> >"The actions taken prior to commit are only tentative (i.e., not
persistent
> >and not visible to other activities).... Commit makes the tentative
actions
> >visible to other transactions.  Abort makes the tentative actions appear
as
> >if the actions never happened."
> >to:
> >"The actions taken by a transaction participant prior to commit are only
> >tentative (and so typically are neither made persistent nor made visible
to
> >other activities)...Commit directs the participants to make the
tentative
> >actions final so they may, for example, be persisted and made visible to
> >other transactions.  Abort directs the participants to make the
tentative
> >actions appear as if the actions never happened."
> >
> >
> mm1: Ian and Ram, have you considered this is introductory, high-
> level text that is more focused on the user point of view rather
> than the implementor?

Yes, this is introductory text that is not focussed specifically on the
implementor.


> The AT Coordinator control whether all
> actions that participants are responsible for succeed or abort,
> rather than directly persist or rollback data.  If the participant
> is a durable one (rather than a volatile one), it could be
> persisting data. Or, do you believe your text is consistent with
> that view? Thanks.

The text I proposed is trying to more precisely state exactly this. Its a
very minor thing but I think that if we are going to clarify ths text as
suggested in issue 102 then we should make it clear that the AT commit is
simply directing the participants to make their tentative changes final.
Previously the text might have been interpreted that "tentative" meant "not
persistent" and so "commit" meant "make persistent". For some participants
that is exactly what it means but not, typically, for Volatile2PC
participants. And, even for those participants where commit does result in
a tentative change being made persistent, it is the participant (or a
resource manager it delegates to) and not the coordinator that deals with
persisting the data and making it visible.