Re: [wsrf] RE: [wsn] Issue: TerminationTime property should allow Set/Update

[Rich Thompson <richt2@us.ibm.com>]

I have been recently trying to catch
up with all the threads/work happening with WSRF and had the following
statement stand out when reading the updated Resource Lifetime spec:

A WS-Resource MUST NOT allow the TerminationTime resource property to be modified by a SetResourceProperties request message as defined in [WS-ResourceProperties].

This raised the question I knew had been discussed in a previous email thread (copied below), which I would rephrase to "Why is this a resource property if the normal resource property operations are disallowed in favor of a specialized operation?". I think this question will be frequently raised by those who have not had the benefit on participating in these discussions. I understand the issues with a potentially chatty protocol if the semantics of SetTerminationTime are followed using only the base resource property operations and the mismatch between the semantics of SetTerminationTime and SetResourceProperties (i.e. set to a value in the future relative to the supplied value vs. set to the supplied value) and propose that a compromise is in order. I suggest the above sentence be changed to:

A WS-Resource SHOULD NOT allow the TerminationTime resource property to be modified by either a SetResourceProperties or UpdateResourceProperties request message as defined in [WS-ResourceProperties].  If a WS-Resource chooses to allow modification of the TerminationTime resource property via the SetResourceProperties or the UpdateResourceProperties request message, the WS-Resource MUST either accept the TerminationTime value specified in the request or fault as normatively described in [WS-ResourceProperties].

This allows the option of treating the resource property in the standard way provided the WS-Resource is willing to abide by the standard semantics with no impact on the semantics defined by wsrf-rl.

Rich Thompson
OASIS WSRP TC Chair

P.S. In rereading this before hitting send, I think I could be easily convinced to drop the first sentence as it then just becomes a conformance requirement on those WS-Resources that allow the TerminationTime property to be changed via a setResourceProperties or UpdateResourceProperties request message. Do people think the spec should be making a recommendation in this area or simply calling out the difference in semantics and the resulting additional conformance issue involved when the base property operations are supported?

Tom Maguire/Hawthorne/IBM@IBMUS

08/13/2004 04:28 PM

To	"Murray, Bryan P." <bryan.murray@hp.com>
cc	wsrf@lists.oasis-open.org
Subject	[wsrf] RE: [wsn] Issue: TerminationTime property should allow Set/Update

Ouch, the semantic of SetResourceProperties is to set the RP to the value
specified.  Would you
say the behaviour was okay for non-monotonic changes or for other resource
properties?  Can a
service just decide to put whatever value it feels like in an RP and return
a non-fault response?

Tom


"Murray, Bryan P." <bryan.murray@hp.com> wrote on 08/13/2004 01:41:25 PM:

> Why would you not allow setting the TerminationTime in the future if you
> use Set/Update? It is entirely possible that the termination time is
> modified from the value the client sent in the request, but so what? If
> the client sets the time 1/2 hour in the future and sends this message
> every 15 minutes - what difference does it make if the client never
> knows that the service is changing this 1/2 hour to 20 minutes?
>
> Bryan
>
> -----Original Message-----
> From: Tom Maguire [mailto:tmaguire@us.ibm.com]
> Sent: Friday, August 13, 2004 8:59 AM
> To: Murray, Bryan P.
> Cc: wsn@lists.oasis-open.org
> Subject: Re: [wsn] Issue: TerminationTime property should allow
> Set/Update
>
>
>
>
>
> So I presume you told the developer that SetTerminationTime MAY set the
> TerminationTime to a time in the future and that then requires the
> return of the new TerminationTime.  This is different behavior from the
> SetResourceProperties operation.  I am intrigued though by the notion of
> allowing SetResourceProperties on TerminationTime.  I guess in this case
> that the WS-Resource would be prohibited from setting the
> TerminationTime to a time in the future or if it wanted to do so it
> would be required to fault.
>
> Tom
>
>
> "Murray, Bryan P." <bryan.murray@hp.com> wrote on 08/13/2004 11:18:39
> AM:
>
> > I was approached by one of our developers regarding the
> > TerminationTime resource property in WS-ResourceLifetime. His comments
> are as follows:
> >
> >
> > There is a "TerminationTime" resource property, in addition to an
> > operation called "SetTerminationTime".
> >
> > To retrieve the value of the current termination time, you can make a
> > "normal" WSRF-RP GetResourceProperty call and ask for the
> > TerminationTime property - as dictated by the WS-RP spec.
> >
> > However, to set a new value, you MUST call the special
> > SetTerminationTime operation - the spec explicitly states, "A
> > WS-Resource MUST NOT allow the TerminationTime resource property to be
>
> > modified by a SetResourceProperties request message as defined by
> > WS-RP".
> >
> > I don't see why the need for this distinction??  It makes implementors
>
> > have to deal with this special case with no apparent need for it.
> >
> > Why can't the WSRF-RL spec rely on the setting capabilities of WSRF-RP
>
> > just as it relies on the getting capabilities of WSRF-RP?  Why does it
>
> > use GetResourceProperty but not SetResourceProperties and instead
> > define its own special operation for setting the property?  Seems like
>
> > there is no real need for this special case - but yet it causes
> > implementors to have to write code for this "one-off" - we have here a
>
> > required Resource Property that we must implement, but yet we must
> > "disable" some of the generic WS-RP functionality for this one
> resource property.
> >
> >
> > Bryan
> >
> > P.S. After I explained the reason for having the SetTerminationTime
> > operation, the developer agreed there was valid reason to have the
> > operation, but still felt there was no reason to have to write special
>
> > code to disable the ability to use Set to modify the property. There
> > will be many clients that do not need to know exacly when the
> > termination time really is and will just periodically renew the
> > subscription until it is no longer needed.
>