mina-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Bernd Fondermann <bf_...@brainlounge.de>
Subject Re: [Vysper] Error stanzas return too much of the original
Date Tue, 14 Jul 2009 07:50:27 GMT
Niklas Gustavsson wrote:
> On Mon, Jul 13, 2009 at 6:19 PM, Bernd Fondermann<bf_jak@brainlounge.de> wrote:
>>> Yeah, I read this as well as also makes the interpretation that we
>>> should include the complete stanza, but examples seems to contradict
>>> that.
>> What kind of contradiction is it?
>> (I must admit, my mental svn up failed when I wanted to get me to HEAD
>> on this discussion.)
> 
> :-)
> 
>> All examples I see are completely without error echos.
>> I didn't find a single example for an error stanza containing the whole
>> or only a part of the original. Can you point me to one?
> 
> The examples I refer to are those in the RFC and XMPP Ping specs. For
> example, in http://www.ietf.org/rfc/rfc3920.txt page 39 you will find
> some example. And in http://xmpp.org/extensions/xep-0199.html, you can
> find examples in each of the use cases (4.X). All of these include the
> original stanza, without the outermost iq element.
> 
>>> For one, Smack can not handle our current error stanzas as showed by
>>> the integration tests I've written for XMPP Ping (not yet committed
>>> since they do not currently pass). In these, when sending an error,
>>> Smack will not be able to locate the <error> element.
>> Why would Smack try and locate it? What happens if it can't? Does it
> 
> Smack provides a IQ.getError() method for retrieving details on the
> error. In interacting with the current Vysper, this will return null.
> With the proposed patch, it will be correctly populated.
> 
>> At xsf.org, there are two lists,
>>  JDev@jabber.org and
>>  Standards@xmpp.org
>> which coul prove helpful if we are in doubt.
> 
> Yeah, that's a good suggestion. At least I think the phrasing in 9.3.1
> should be clarified. I'll ask over there. Given the spec examples
> however, I think we should change the Vysper implementation for now.

Now I get it. You are right. Sorry for the delay in comprehension.
It makes sense to not echo the outer stanza frame because the reference
is already clear from the answer. However, if the malicious part is in
the 'frame', why the error is returned is not clear without knowing the
original. But maybe these kinds of errors would better result in a
stream error, I don't know.

+1 to apply the proposed patch.

Thanks for your patience,

  Bernd




Mime
View raw message