> -----Original Message-----
> From: Pete Cordell [mailto:[EMAIL PROTECTED]
> Sent: Tuesday, December 09, 2008 8:15 AM
>
> > Malformated INFOs should be hanlde like any other malformed SIP method.
>
> But only the application (and not the SIP layer) knows whether it is
> malformed or not.  So the SIP layer needs to ask the application whether
> it
> will accept the message before giving a response.

That's not necessarily true.  I could have a firewall-type checker in the SIP 
layer that verifies XML syntax is valid, which knows nothing about the 
semantics of the elements/attributes themselves, before passing it on to an app 
layer.

And before anyone says "but we can't require it to do so and we have to 
accommodate a layered model only understanding the content in the app-layer" - 
that's NOT the issue.  Of course we have to allow for only the app-layer 
understanding content - we should even *assume* it.

But the reality is that we already allow a 4xx response to be sent for a bunch 
of sip-layer error reasons.  Because of that, the app-layer *sender* of an INFO 
request has to accommodate "delivery-failure" based on a 4xx response already.  
Period.  Therefore, an _integrated_ model *receiver* would simply send one of 
those 4xx responses if it failed content formatting error checks when it 
receives an INFO request, because from a functional app-layer perspective to 
the INFO *sender* it is no different than a delivery-failure.  There is nothing 
we can do to stop that behavior, because there's nothing broken with that 
behavior, either in terms of interop or semantics.  We can try to say "even if 
the content is malformed you MUST send back 200 and then a new INFO with an 
error code", but that will be summarily ignored.

So the question is simply if we want a specific response code for content 
formatting errors, such as 488, to help troubleshooting - or if they should 
just send 400 or 415 as they do today.  Again, this doesn't prevent a fully 
layered model from sending 200 back and then some INFO request upstream with an 
error code - it just makes it easier to troubleshoot what an integrated model 
can already do by sending back a 400 or 415.

-hadriel

_______________________________________________
Sip mailing list  https://www.ietf.org/mailman/listinfo/sip
This list is for NEW development of the core SIP Protocol
Use [EMAIL PROTECTED] for questions on current sip
Use [EMAIL PROTECTED] for new developments on the application of sip

Reply via email to