[Standards-JIG] JEP 60: Most Recent Items defined?

Peter Saint-Andre stpeter at jabber.org
Fri May 20 22:05:54 UTC 2005


On Fri, May 20, 2005 at 03:05:43PM -0700, Chris Mullins wrote:
> PSA Wrote:
> > Do you feel that the JEP needs to specify this?
> 
> I believe it needs to. 
> 
> The reason is one of compliance testing. I have tests that publish items
> to the server and then request items back and verify the correct items
> are returned. 
> 
> Without having the JEP specify the order, there's no way to write these
> tests, and therefore no mechanism by which to test compliance. The tests
> might pass on the SoapBox PubSub implementation, but fail on the others
> - yet both would be in compliance with the spec. Clearly, this is not
> ideal.
> 
> In the example that I originally used:
> > <iq [...]>
> >     <pubsub xmlns="http://jabber.org/protocol/pubsub">
> >         <publish node="somenode">
> > 
> >             <item id="left">
> >                 <leftnode/>                 
> >             </item>
> > 
> >             <item id="right">
> >                 <rightnode/>
> >             </item>
> >         </publish>
> >     </pubsub>
> > </iq>
> 
> When the test publishes these two items, then requests the "most recent
> item", the test is useless as a compliance test if sometimes "leftnode"
> is correct, and sometimes "rightnode" is correct. Specifying this makes
> these cases testable. 

Well, if the JEP specifies this, then it specifies how an implementation
must parse XML nodes, and that seems problematic to me -- especially
since it is possible that the order of child elements could be changed
en route (e.g., by an intermediate router).

Naturally the compliance test could simply say that the concept of "most
recent item" is ambiguous in these cases, no?

/psa






More information about the Standards mailing list