Re: Fwd: Re: SCTP mapping to data channels

Harald Alvestrand wrote:
> This way, reliable, sequenced datagrams would be on one channel;
> unreliable, unsequenced ones would be on another, giving consistency.

This is entirely at the discretion of the sender, right? If the receiver 
needs to rely on a particular data channel being ordered or reliable, 
it's probably making a number of assumptions about what's in it (that 
the sender would do well to conform to if it wants the receiver to work 
correctly).

As long as we don't expose whether or not a data channel is reliable in 
the API on the receiving side (and I don't see a need to... the sender 
can drop or reorder datagrams before it ever gives them to the SCTP 
stack, so it's not like we're exposing the receiver to anything it 
wouldn't have had to deal with before), we can still make the API on the 
sending side use one set of options per channel without imposing any 
additional restrictions on what non-browser peers might do. That would 
give us the option to add a sender API with per-packet options later if 
we ever saw the need.

Received on Monday, 6 February 2012 22:20:42 UTC