Re: [rtcweb] #13: Transport of DATA_CHANNEL_OPEN

[Harald Alvestrand <[e-mail]>]

On 04/18/2013 12:01 AM, Martin Thomson wrote:
> On 17 April 2013 14:58, Harald Alvestrand <[e-mail]> wrote:
>> I think I (on behalf of applications I'll be writing) am in the camp of "If
>> I asked to be told what the incoming data channel was (by not registering to
>> handle data without an OPEN), and didn't get told what it was, I don't want
>> the data".
>>
>> Perhaps provide a few buffers to handle packet reordering.
>> But beyond that: Throw the data away.
> What if the data is ordered?  Couldn't the data you actually get
> depend on the data that got discarded?
If the data is required to be delivered in-order, it can't be delivered 
before the (in-order) OPEN packet I'm waiting for anyway, so the problem 
cannot occur.

The problem can occur if data can be sent reliably, but without ordering.

I'm willing to consider the option of saying that "throw the data away" 
means "report to the sender that this data channel closed with an 
error". It's even reasonably correct; what the sender expected to happen 
did not happen.

(The receiver will never know that a data channel was attempted. It 
never opened.)