Re: DataChannels API

On 9/1/2012 5:19 AM, Timothy B. Terriberry wrote:
> Randell Jesup wrote:
>> Is the single "reliable" attribute enough?  Do we need to expose the
>
> I would think we would want a separate "ordered" attribute as well. 
> Guaranteed/non-duplicated delivery and ordered delivery are orthogonal 
> concepts in SCTP, and I don't see why we wouldn't want to expose that.

Agreed.  I don't see a ton of downside to exposing it, though if we 
wanted to expose all the parameters as a copy of the input dictionary 
object that's more work, as opposed to exposing some normal readonly 
attributes.  In theory they could be settable, but that would complexify 
the protocol with little gain, since you can just open a new channel and 
close the old one.

So: expose ordered.  Open question: expose values for rexmit num and time?

-- 
Randell Jesup
randell-ietf@jesup.org

Received on Sunday, 2 September 2012 02:19:10 UTC