Do we restrict subscriptions to exactly one topic? I thought wildcards
were supported. There is no conceptual reason
getCurrentMessage couldn't support wildcards, but the semantics would
require a little care. Personally I'd rather not go there without good
reason.
I didn't see (4) as equivalent to "topic not supported". Topic not
supported is in there, but it's conjoined with "exactly one", which I
took as primary.
Steve Graham wrote:
David:
Thanks for examining the WS-BaseN
draft.
With regards 2.17:
>2.17 Calls for three faults (InvalidTopicExpressionFault,
TopicNotSupportedFault, TopicPathDialectUnknownFault) dealing with bad
topic expressions >to be shared by both Subscribe and
GetCurrentMessage.
The text mentions invalid topic expression and unknown topic path
dialect, but not with these >particular names. It doesn't mention
TopicNotSupportedFault.
Subscribe lists the following faults
(I number them in this email for ease of reference):
- · Unable
to create subscription
- · TopicPath
Dialect not understood
- · Invalid
TopicExpression
- · Given
TopicExpression did not match exactly one Topic supported by the
NotificationProducer
- · The
Topic is supported but there is no NotificationMessage currently
available
on that Topic
- · The
Topic does not currently have a message associated with it.
- · Others,
TBD
2 corresponds TopicPathDialectUnknownFault,
3 corresponds to InvalidTopicExpressionFault
and 4 corresponds to TopicNotSupportedFault
GetCurrentMessage Lists the
following
faults:
- · TopicPath
Dialect not understood
- · Invalid
TopicExpression
- · Given
TopicExpression did not match exactly one Topic supported by the
NotificationProducer
- · The
Topic is supported but there is no NotificationMessage currently
available
on that Topic
- · Others,
TBD
1 corresponds TopicPathDialectUnknownFault,
2 corresponds to InvalidTopicExpressionFault
and 3 corresponds to TopicNotSupportedFault.
Furthermore the WSDL expresses the
convergence
even more strongly:
<wsdl:operation
name="Subscribe">
<wsdl:input
message="wsntw:SubscribeRequest" />
<wsdl:output
message="wsntw:SubscribeResponse" />
<wsdl:fault
name="ResourceUnknownFault"
message="wsntw:ResourceUnknownFault"
/>
<wsdl:fault
name="TopicPathDialectUnknownFault"
message="wsntw:TopicPathDialectUnknownFault"/>
<wsdl:fault
name="InvalidTopicExpressionFault"
message="wsntw:InvalidTopicExpressionFault"
/>
<wsdl:fault
name="TopicNotSupportedFault"
message="wsntw:TopicNotSupportedFault"
/>
<wsdl:fault
name="SubscribeCreationFailedFault"
message="wsntw:SubscribeCreationFailedFault"/>
</wsdl:operation>
<wsdl:operation
name="GetCurrentMessage">
<wsdl:input
message="wsntw:GetCurrentMessageRequest"/>
<wsdl:output
message="wsntw:GetCurrentMessageResponse"/>
<wsdl:fault
name="ResourceUnknownFault"
message="wsntw:ResourceUnknownFault"
/>
<wsdl:fault
name="TopicPathDialectUnknownFault"
message="wsntw:TopicPathDialectUnknownFault"/>
<wsdl:fault
name="InvalidTopicExpressionFault"
message="wsntw:InvalidTopicExpressionFault"
/>
<wsdl:fault
name="TopicNotSupportedFault"
message="wsntw:TopicNotSupportedFault"
/>
<wsdl:fault
name="NoCurrentMessageOnTopicFault"
message="wsntw:NoCurrentMessageOnTopicFault"
/>
</wsdl:operation>
</wsdl:portType>
>Further, the list for
Subscribe
includes faults that only apply to getCurrentMessage (topic expression
names more than one topic, no message available >for topic).
Hmmm...
· The
Topic is supported but there is no NotificationMessage currently
available
on that Topic
· The
Topic does not currently have a message associated with it.
Should not appear with
Subscribe.
This appears to be a stupid copy/paste error on my part. Good
catch.
> Unless I've grossly
misunderstood,
the issue should remain open.
The other possibility is that I
misunderstood
the proposed solution to the issue, in which case we should discuss at
the next telecon.
sgg
++++++++
Steve Graham
(919)254-0615 (T/L 444)
STSM, On Demand Architecture
Member, IBM Academy of Technology
<Soli Deo Gloria/>
++++++++
Summary: All but 2.17 appear to be closed. In detail:
- 2.3 description might be a little clearer that
the computed
"absolute time" can cause a fault even if "duration"
was specified, but basically good. Closed, but I may tweak language
in next rev.
<>
- 2.11 SubscriptionReference added to Notify, with
explanatory
text. Closed.
- 2.17 Calls for three faults (InvalidTopicExpressionFault,
TopicNotSupportedFault, TopicPathDialectUnknownFault) dealing with bad
topic expressions to be shared by both Subscribe and GetCurrentMessage.
The text mentions invalid topic expression and unknown topic path
dialect, but not with these particular names. It doesn't mention
TopicNotSupportedFault. Further, the list for Subscribe includes
faults that only apply to getCurrentMessage (topic expression names
more
than one topic, no message available for topic). Unless I've grossly
misunderstood, the issue should remain open.
- 2.18 The
SubscriptionManager section
now RECOMMENDs WSRL but SMs MAY support other means of destruction.
They
MUST advertise what they support. Closed.
- 2.19 The NP may now
return an "Unable
to subscribe" fault, for whatever reason. Closed.
- 2.21 Clarifications of
schema have
been incorporated. Closed.
- 2.22
{Pause,Resume}Subscription[Response]
are all explicitly defined as {any}. Closed.
- 2.29 TerminationTime is
now an
optional element in SubscribeResponse. Closed
- 5.2 Description of
pseudo-schemas
follows Steve's email, with explicit mention that it is "similar
to the WSDL 2.0 Part 1 specification [WSDL 2.0". We
may tweak this later if a more official description appears, but for
now
the issue is Closed.
Steve Graham wrote:
The latest version of WS-BaseN [1]
addresses:
WSN 2.3, 2.11, 2.17, 2.18, 2.19, 2.21,
2.22,
2.29, 5.2
Would someone volunteer to verify
the
issues were properly executed?
I am now going to start diving into the issues resolved at the Oct f2f.
[1]http://www.oasis-open.org/apps/org/workgroup/wsn/download.php/10147/wsn-WS-BaseNotification-1.3-draft-01a.doc
sgg
++++++++
Steve Graham
(919)254-0615 (T/L 444)
STSM, On Demand Architecture
Member, IBM Academy of Technology
<Soli Deo Gloria/>
++++++++
|