NEW Issue: WS-C: Clarify subordinated CreateCoordinationContextbehaviour

[Alastair Green <alastair.green@choreology.com>]

Issue name -- WS-C: Clarify subordinated CreateCoordinationContext behaviour

PLEASE DO NOT REPLY TO THIS EMAIL OR START A DISCUSSISON THREAD UNTIL 
THE ISSUE IS ASSIGNED A NUMBER.

The issues coordinators will notify the list when that has occurred.

Target document and draft:

Protocol:  Coord

Artifact:  spec

Draft:

Coord spec working draft uploaded 2005-12-02

Link to the document referenced:

http://www.oasis-open.org/committees/download.php/15738/WS-Coordination-2005-11-22.pdf

Section and PDF line number:

Section 3 "Coordination Service", ll. 181-209


Issue type:

Design / editorial


Related issues:

None


Issue Description:

Precisely specify normative behaviour of CreateCoordinationContext 
against existing (superior) context


Issue Details:

[This issue stems from Choreology Contribution issue TX-17.]

There are two intertwined problems. One is presentational, the other 
more substantive. The initial part of Section 3 (up to and inluding the 
diagram, Figure 2., on p. 10) uses an example to convey three things:

    a) the basic interactions of a registering application service and a 
Registration Service, and their
    relationship to Coordinators and Participants
    
    b) the notion of an interposed Coordinator, and
    
    c) some seemingly normative rules relating to how interposed 
coordination works.
 
It should be made clear what is exemplification, and what is normative. 
It is possible that the exposition of
basic (C-P) and interposed (C-P/C-P) behaviours may be easier to grasp 
with two successive and related
examples. These are primarily editorial matters. In the proposed 
resolution I will call this sub-issue a).

The substantive problem concerns the implied or stated normative rules 
of interposed coordination. There are
two statements, which are too restrictive -- or perhaps which seem too 
restrictive because they are actually
intended only as examples of one possible use of the protocol, but 
appear to have normative weight because
there are no other statements at all in the specification about the 
behaviours being discussed.

These statements are contained within the following three call-outs 
which refer to the diagram. They are  flagged with the inserted tags 
[Statement 1] and [Statement 2].

ll. 197-206

    "3. App2 prefers CoordinatorB, so it uses CreateCoordinationContext 
with Ca as an input to interpose
    CoordinatorB. CoordinatorB creates its own CoordinationContext Cb 
that [Statement 1] contains the same
    activity identifier and coordination type as Ca but with its own 
Registration service RSb.

    "4. App2 determines the coordination protocols supported by the 
coordination type Q and then Registers for
    a coordination protocol Y at CoordinatorB, exchanging Endpoint 
References for App2 and the protocol
    service Yb. This forms a logical connection between these Endpoint 
References that the protocol Y can use.

    "5. [Statement 2] This registration causes CoordinatorB to forward 
the registration onto CoordinatorA's
    Registration service RSa, exchanging Endpoint References for Yb and 
the protocol service Ya. This forms a
    logical connection between these Endpoint References that the 
protocol Y can use."

Dealing with Statement [1] first:

It is wrong to mandate, in the WS-Coordination specification, that the 
activity identifier of the newly
created CoordinationContext should be the same as that of the 
pre-existing (superior) Coordinator's
CoordinationContext. (Proposed resolution sub-issue b).)

For the purposes of correct operation of the interposed coordinator per 
se, the identity of the top transaction and the identity of the bottom 
transaction are irrelevant.

For the purposes of data sharing, every data-accessing participant in a 
tree or sub-tree that is known to have
an atomic outcome (e.g. WS-AT and WS-BA AO) should have available an 
identical value which can be used as an
accessor token. A single, tree-wide value for all propagated context's 
/Identifier is a convenient way of
achieving this. It would be appropriate to insert a rule in WS-AT and 
WS-BA to the effect that a new atomic
context which is registered against an existing /atomic/ outcome must 
inherit the /Identifier value of the
existing context. It is inappropriate to place this rule in WS-Coord.

(The above point implicitly rests on the assumption that, for atomic 
outcomes, every participant which
registers with an RS associated with a given /Identifier will receive 
the same outcome. Issues are required to
explicitly state this for WS-AT and WS-BA.)

To create a mixed outcome, one must either send differently identified 
contexts, or send identically
identified contexts with a flag indicating that they cannot be relied 
upon to receive the same outcome (the
MixedOutcome flag in a WS-BA context permits the latter option).

In this case we cannot assume data sharing, but nor can we rule it out 
(it may still be useful to have a
common accessor identity for more sophisticated forms of isolation 
control than outright blocking). At this
point, we might conclude that the identifier-inheritance rule is 
strictly unnecessary but harmless.

Scenario: A BA tree which combines MixedOutcome with AtomicOutcome. 
(Other cases can be created involving
combining WS-AT and WS-BA, and we have no control (at the level of 
WS-Coordination) on future interposition or
sub-coordination behaviours for entirely new coordination types.)

A consuming service CS-Buyer uses three provider services: PS-Goods, 
PS-Shipping, PS-Insurance. There are two
viable (successful) outcomes possible -- {G,S}, {G,S,I} -- so we use 
MixedOutcome to govern the relationship
of the CS to the three PS. Inside each PS (and these are likely to be 
separate legal entities), there may be
several services that are composed to offer a single transactional, 
contingent service. To take one example:
PS-Goods uses internal services IS-Stock, IS-Credit, IS-Accounting to 
create a single atomic outcome
"OrderGoods" operation. PS-Goods creates a WS-BA AtomicOutcome 
sub-coordinator, and hooks it in to the CS-
propagated B2B transaction.  

Imagine that IS-Credit accesses an external bureau service. PS-Insurance 
uses the same service. Both Goods and
Insurance use the CS-Buyer transaction id to identify access to the 
credit system. If this has access control
implications (causes data sharing) then the fact that Insurance may fail 
independently of Goods can cause
unexpected side effects, if the rule "common id = atomic outcome" is 
relied upon.
 
(Less strained examples can be created using trees which combine WS-BA 
MO and WS-AT in intra-enterprise environments where data sharing is more 
likely.)

WS-Coordination should say nothing about identifier inheritance. It is a 
matter for coordination protocols,
for transaction domain bridging software, and for applications (which 
may use identifiers to create groupings
or dependencies which are far more subtle).

Much of the above discussion hinged on the business need to combine 
coordination types in a transaction tree.
But the second part of [Statement 1] implies that interposition can only 
occur between contexts of identical
coordination types. As Tom Freund pointed out at the first F2F, the 
design intent hitherto has been to permit
combined-type tree creation. (Proposed resolution sub-issue c).)

Turning to Statement [2] (proposed resolution sub-issue d)):

The practice of having a subcoordinator delay registration with its 
nominated superior coordinator until it
first receives registration(s) from Participant(s) is viable, given 
certain rules. So is the practice of
eagerly registering subcoordinators as they are created. The lazy/eager 
choice is not the property of WS-
Coordination, and the spec should make it clear that subcoordinator 
registration time is not defined.


Proposed Resolution:

Sub-issue a). Editorial work required to clarify example, and 
accompanying figure, and to create a section that describes the model 
and normative operation of sub-coordination.

Sub-issue b). Remove any normative reference to subcoordinator 
identifier inheritance in WS-Coordination.

Sub-issue c). Remove any normative statement relating to sameness of 
superior and sub-coordinator coordination
types or protocols.

Sub-issue d). Remove any normative statement relating to the timing of 
sub-coordinator registration, other
than to state that the sub-coordinator must be registered to take part 
in activity completion.