Re: [rtcweb] Unsupported PPID (Re: draft-ietf-rtcweb-data-channel failure handling description)

[Michael Tuexen <[e-mail]>]

On 27 Aug 2014, at 13:16, Harald Alvestrand <[e-mail]> wrote:

> On 08/26/2014 11:22 PM, Martin Thomson wrote:
>> On 26 August 2014 13:36, Michael Tuexen
>> <[e-mail]> wrote:
>>> What I don't understand how ignoring gives you extensibility. If one sides sends a new PPID, and
>>> the other sides ignores it, the sender (at the SCTP level) thinks that the message was received
>>> without any problem, since SCTP doesn't take the PPID into account. So user messages are lost.
>> I think that the use of a new PPID can be negotiated prior to use, so
>> there aren't big problems with any action recommended here.  Resetting
>> the stream might work, assuming that we can distinguish between an
>> error and a graceful close correctly, that is.
> 
> If there is negotiation, there is no problem.
> 
> If there is no negotiation, starting to use the new feature will either lead to ignored messages or lead to a crashed connection.
> 
> Ignored messages are not a problem if they form part of an ignorable extension (for instance if someone decided to once in a while send checksums over all the data sent so far - doesn't matter to those that don't understand it, since it doesn't carry any extra data).
> 
> Crashed connections are usually a problem, so that will mean "no unnegotiated extensions, ever".
> 
> I'm happy with either way of doing it, as long as we're explicit about it.
I agree with you, don't have a preference. Up to now the consensus was to close the channel...
I can add a sentence to make clear that this means no unnegotiated  extensions.

Best regards
Michael
> 
> 
> 
>