[asterisk-dev] [Code Review]: Ensure that CDRs for a caller in a Queue who is not answered is NO ANSWER

[Matt Jordan]

> On Aug. 9, 2012, 3:06 p.m., Matt Jordan wrote:
> > Bumping the review.
> 
> Tilghman Lesher wrote:
>     The problem is that you're only looking at the simplest case.  Suppose a call comes in, they talk to a person, and then they get transferred to a Queue, and they hang up before the Queue is answered (in any of the scenarios above).  Your essentially saying that the billsecs up until that time don't count and that the CDR should therefore both indicate "NO ANSWER", which is patently false, as well as recording some billsecs, because the call was, in fact, answered.  That will confuse all sorts of billing systems, and thus is really of the type of fix that shouldn't go in unless you account for problems like this; well, no, shouldn't go in, period.  There are too many call scenarios where trying to manipulate this value after the call was answered will be problematic and will break billing systems.  In other words, don't touch it.
> 
> Paul Belanger wrote:
>     This was the reason for my previous comment.  I was under the impression that we no longer did any CDR fixes, and recommended people use CEL, simple because we could never ensure it would not break other scenarios.

The assumption that we 'no longer did any CDR fixes' is incorrect.  CDRs have not been deprecated.  Would I like to deprecate them?  Sure - but I don't think that's a valid option at this time.  If I'm wrong in that then by all means, send an e-mail to the -users list telling everyone that CDRs are deprecated and see what response you get.

What we aren't going to attempt to do is define behavior that cannot be defined given the limitations imposed by CDRs.

So what behavior is undefined?  Well, one area that is certainly undefined in CDRs is what happens in transfer scenarios.  We've never claimed that the resulting CDR record in a blind or attended transfer is going to account for all of the actions that happen between the various channels.

So let's look at the case of an attended transfer as Tilghman outlined without this patch:

Phone A calls Phone B.  Phone B transfers Phone A into a Queue with Phone C as a member (unpaused) and Phone D as a member (paused).  Phone A hangs up.

Do you get a billsec value?  Yup.  Do you get a disposition?  Yup - BUSY in this case.  That hardly makes any sense, but - of course - CDR behavior in transfer scenarios is undefined.

Lets try this one.  Phone A calls Phone B.  Phone B transfers Phone A into a Queue with Phone C as a member and Phone D as a member, both unpaused.

Now its even worse: if there's an Answer() before the Queue, you'll get a CDR with a disposition of ANSWERED - if there isn't an Answer() before the Queue, you get a disposition of NOT ANSWERED.  In both cases, however, you still have a billsec value.

So Tilghman's problem already exists without this patch.

I'm not claiming to fix the case of transfers, nor am I trying to 'fix' all issues with CDRs.  In this case, however, the simple behavior of CDRs in Queue with multiple agents has multiple behaviors, none of which are predictable.  I'm not sure how this patch doesn't at least make the simpler case consistent.  Since that behavior has been defined with respect to a parallel Dial, matching that behavior with Queue makes some sense.


- Matt


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/2064/#review6895
-----------------------------------------------------------


On July 25, 2012, 4:54 p.m., Matt Jordan wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/2064/
> -----------------------------------------------------------
> 
> (Updated July 25, 2012, 4:54 p.m.)
> 
> 
> Review request for Asterisk Developers.
> 
> 
> Summary
> -------
> 
> When a caller enters a queue and no queue member answers the call, some odd behavior can be reporter in the resulting CDR records, depending on the state of certain queue members.
> 
> * If no queue members are paused, the CDR result is whatever the result was prior to going into app_queue.  If the call was answered, this is ANSWERED; otherwise, it is NO ANSWER.
> * If some queue members are paused, the CDR result is BUSY.  This patch changes this to NO ANSWER.
> * If all queue members are paused, the CDR result is again whatever the result was prior to going into app_queue.
> * If the caller hangs up, times out, or presses '*' with the 'h' option, the result is again not set.
> 
> Similar to app_dial, if the caller in the queue does not get answered, we should set the CDR record appropriately.  In general, for the scenarios listed above, this is NO ANSWER.
> 
> This patch is a only slightly modified version of the one supplied by the issue reporter.
> 
> Since this is arguably a change in behavior, some tests were written to cover portions of this behavior in https://reviewboard.asterisk.org/r/2063.
> 
> 
> This addresses bug AST-906.
>     https://issues.asterisk.org/jira/browse/AST-906
> 
> 
> Diffs
> -----
> 
>   /branches/1.8/apps/app_queue.c 370418 
> 
> Diff: https://reviewboard.asterisk.org/r/2064/diff
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Matt
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20120809/de948e24/attachment-0001.htm>