Not included at all? Interesting. [<lc>Taking from David Ehnebuske's
"Principle of Least Astonishment" - it seems entirely reasonable to say
nothing about a deletion in the future of an entity for a period that
preceeded its deletion. </lc>]
When I was trying to answer the question I re-read the spec, and it seemed
that an entry which changed is supposed to be included in the results in its
current state, but if it was consequently deleted, one cannot report its current
state. The spec does not state what to do in this case - that's what I think
should be clarified. [<lc> what section do you
suggest </lc>]
My belief that it should be clarified is strengthened. After all:
1. my interpretation is that it should be reported as deleted
2. your interpretation is that it should not be reported [<lc>back
to David's principle: there's nothing to be said about an entity that has
yet to be deleted. Futhermore, I think that the case of the virtual delete is
fairly clear also. </lc>]
Sounds like a potentially nasty interop issue.
-----Original Message----- From: Luc Clement
[mailto:luc.clement@systinet.com] Sent: Tue 13-Sep-05 7:23
To: Rogers, Tony; uddi-spec@lists.oasis-open.org Cc:
Subject: RE: [uddi-spec] Anomaly in
get_subscriptionResults
--- ooops you must think: surely he's lost his mind... I meant
to reply to this email not the notification of postponement of the
FTF.
We've interpreted and implemented as follows: the entity deleted
on 1May 2005 IS NOT included in a result of get_subscriptionResults over
time period of 1Jan 2004 to 1 Jan 2005.
This is consistent with the
intent of the spec. What specifically in the spec causes you to believe
this to be clarified (too lazy to go hunt
for it).
Luc
-----Original Message----- From:
Tony.Rogers@ca.com [mailto:Tony.Rogers@ca.com] Sent:
Thursday, September 08, 2005 21:52 To:
uddi-spec@lists.oasis-open.org Subject: [uddi-spec] Anomaly in
get_subscriptionResults
I was just asked a question, and I think the
answer implies a need for additional clarification of the spec.
If a
user calls the get_subscriptionResults API, specifying a period of, say, 1
January 2004 to 1 January 2005, and a subscription filter that matches a
particular business entity that changed on 1 July 2004, then we should
return the current state of that business entity - I think we're all agreed
on that. But if that business entity was deleted on 1 May 2005, which is
AFTER the requested period, should we return the business as deleted (in
other words, should we include its key in a delete key bag?)?
As far
as I can see, that's what we obliged to do. For contrast, returning the
last state of the business before it was deleted is misleading, and
returning the state of the business as at the end of the period is
explicitly excluded by the spec.
This means that the deleted bag
contains not just those entities which were deleted (real or virtual)
during the period specified, but ALSO those which were deleted (but only
REAL deletes, not virtual ones) since.
Right?
I think we
should clarify this in the spec, because it's an anomalous and perhaps
unexpected result.
|